Skip to content

Open-source agentic AI Search Execution Platform. Track AI visibility, find opportunities, and let agents turn signals into prioritized actions and drafted fixes across ChatGPT, Claude, Gemini, Google AI Overviews, Google AI Mode, Perplexity, Grok, and Microsoft Copilot.

docs(help): add the Ansvisor Help knowledge base and its first 56 articles (#1792)

Cihan8184fc527001
diff --git a/.github/PULL_REQUEST_TEMPLATE.md b/.github/PULL_REQUEST_TEMPLATE.md
index dbb1e17f8..b61f59f30 100644
--- a/.github/PULL_REQUEST_TEMPLATE.md
+++ b/.github/PULL_REQUEST_TEMPLATE.md
@@ -27,3 +27,4 @@
- [ ] `yarn typecheck` passes (run from `web/`)
- [ ] `yarn format:check` passes (run from `web/`)
- [ ] Changes are focused — one concern per PR
+- [ ] Help article updated under `help/` (or no user-visible change) — see [help/README.md](https://github.com/ansvisor/ansvisor/blob/main/help/README.md)
diff --git a/CONTRIBUTING.md b/CONTRIBUTING.md
index 8657c8b04..95f6158c8 100644
--- a/CONTRIBUTING.md
+++ b/CONTRIBUTING.md
@@ -24,6 +24,7 @@ ansvisor/
├── web/ # Next.js 16 frontend (TypeScript)
├── server/ # Express backend (Node.js ESM)
├── supabase/ # Database migrations and config
+├── help/ # Product Help articles Chat answers from (help/README.md)
├── scripts/ # Version management tooling
├── docker-compose.yml # Containerized deployment
├── CHANGELOG.md
@@ -173,14 +174,20 @@ We follow [GitHub Flow](https://docs.github.com/en/get-started/using-github/gith
`server/src/lib/demo/index.js`, so every demo brand is rebuilt on its next
visit.
-5. Push and open a pull request against `main`.
+5. If your change alters what a user sees or does, add or update its Help
+ article under `help/` in the same pull request ("Help article updated"):
+ one article per page, feature, setting and main task, in the format of
+ [help/README.md](help/README.md). `yarn test` (run from `web/`) checks
+ the format. A new article starts as `approved: false`.
-6. In your PR description, include:
+6. Push and open a pull request against `main`.
+
+7. In your PR description, include:
- **What** changed and **why**
- **How to test** the change
- Screenshots if there are UI changes
-7. Wait for review, address feedback, and your PR will be merged.
+8. Wait for review, address feedback, and your PR will be merged.
### Checking for schema drift
diff --git a/help/README.md b/help/README.md
new file mode 100644
index 000000000..15fc42d95
--- /dev/null
+++ b/help/README.md
@@ -0,0 +1,41 @@
+# Ansvisor Help
+
+The product knowledge base Chat answers how-to questions from (CHAT-19 §4): one article per page, feature, metric, setting and main task of the live app. Chat answers only from articles with `approved: true`, one short section at a time, and opens the article's `route` with a page card.
+
+## Where
+
+`help/{area}/{slug}.md`, one folder per area: `getting-started`, `visibility`, `prompts`, `mentions`, `citations`, `competitors`, `content`, `site-audit`, `action-center`, `ai-traffic`, `reports`, `settings`. The slug is the file name, in kebab-case.
+
+## Front matter
+
+```yaml
+---
+title: Add prompts to track
+route: /dashboard/prompts/tracked
+keywords: [add prompt, bulk import, track]
+synonyms: [new query, add questions]
+plan: all
+updated: 2026-10-29
+last_checked: 2026-10-29
+owner: Cihan
+app_version: 0.3.0
+approved: false
+sources: [docs/guides/prompts.mdx]
+---
+```
+
+- `route`: the page the article is about, an address of the app (Settings sections by their name, `/dashboard/settings/team`).
+- `keywords` (3 or more) and `synonyms`: the words people ask with.
+- `plan`: the lowest plan the feature needs (`all`, `starter`, `growth`, `enterprise`). Plan limits and prices are never copied into an article: Chat reads them from the plan facts.
+- `updated` / `last_checked`: when it last changed, and when it was last checked against the app; `app_version` the version it was checked against.
+- `approved`: `false` until the owner approves it.
+- `sources`: the documentation it was written from.
+
+## Body
+
+- A one-paragraph intro, then `##` sections. Every section, the intro included, is 150 words or fewer, so it can be answered from alone.
+- Plain English, as the app says it: name pages, tabs and buttons by their labels (**Tracked Prompts**, **Add prompt**), never by their addresses.
+- No code, links, HTML, internal names, database or API details (CHAT-19 §3). Chat adds the page card from `route`.
+- Only what the app does today. When a feature changes, its article changes in the same pull request ("Help article updated").
+
+`web/src/lib/help/validate.test.ts` checks every article against these rules.
diff --git a/help/action-center/action-center-actions.md b/help/action-center/action-center-actions.md
new file mode 100644
index 000000000..0b0debb7d
--- /dev/null
+++ b/help/action-center/action-center-actions.md
@@ -0,0 +1,75 @@
+---
+title: Work with Action Center actions
+route: /dashboard/action-center/actions
+keywords: [actions, action status, assign action, dismiss action, export actions, top actions]
+synonyms: [recommendations, work items, to-dos, action list]
+plan: all
+updated: 2026-10-29
+last_checked: 2026-10-29
+owner: Cihan
+app_version: 0.3.0
+approved: false
+sources: [docs/guides/action-center.mdx]
+---
+An action is one piece of work with an outcome, built from the evidence behind it. Ansvisor groups many signals into one action: six valuable pages with no AI traffic become one "Capture AI traffic" action with six linked signals, not six rows. New actions are added after the daily tracking run, on the **Actions** tab.
+
+## Action categories
+
+- **Growth**: ground you could win, such as demand, topics, platforms or pages where you could appear more.
+- **Protect**: ground you hold that is starting to slip.
+- **Recover**: ground you held and lost, in visibility, citations, mentions or traffic.
+- **Fix**: issues on your own site, such as low audit scores, blocked crawlers or broken links AI sends people to.
+- **Compete**: places where competitors lead you.
+
+## Cards and filters
+
+The cards at the top count and filter the list. Click one to narrow the table, and click it again to clear it.
+
+- **Top Actions**: the five highest-priority open actions.
+- **At Risk**: open Recover and Fix actions.
+- **Opportunities**: open Growth actions.
+- **In Progress**: actions someone is working on.
+- **Highest Priority**: the highest level among open actions.
+
+Below them, filter by category, **Priority**, **Status** and **Assignee**, or search. **Sort** by Priority (the default), Newest, Due date or Recently updated. Filters are kept in the page's address, so a filtered view can be shared. For an online shop, the **From Products** chip lists only the actions sent from Products.
+
+## Status, dismiss and delete
+
+An action's status is **New**, **In Progress**, **On Hold**, **Completed** or **Dismissed**. The row menu sets In Progress, Completed or Dismiss without opening the action.
+
+Dismissing asks why: it doesn't matter to us, it's already done, the data behind it is wrong, or not now. You can also dismiss without saying why. History shows the reason.
+
+**Delete**, for admins and managers only, removes an action for everyone, and it won't come back on its own the next day.
+
+## Inside an action
+
+Click a row to open the action. Under its title you set **Status**, **Assignee** and **Due date**. Its tabs:
+
+- **Overview**: **Why this priority?**, **Why this action?** with the signals behind it, the **Goal** (what done looks like, with a date) and the linked **KPIs**.
+- **Affected**: the prompts, pages, topics, competitors or platforms it is about, with each prompt's demand.
+- **Tasks**: the work, step by step.
+- **Signals**: the evidence. Click one to open it on the Signals tab.
+- **History**: every change, from status and assignee changes to tasks updated or sent.
+
+## Assign work
+
+An action with no one on it belongs to **Everyone** in the workspace until someone takes it. Pick a teammate in the action's **Assignee** field, and set a **Due date** if it has one. The **Assignee** filter on the Actions tab shows one person's actions, or **Everyone** for the unassigned ones.
+
+## Export actions
+
+**Export CSV** on the Actions tab downloads the actions in the current view, as filtered, with their priority, status, assignee, due date, task progress and affected items.
+
+Inside an action, **Copy Data** gives the whole action as Markdown to paste into a doc or ticket, and **Export** downloads it as **Export CSV** or **Export Excel**. The Excel file has one sheet each for the action, its affected items, tasks, signals and history.
+
+## Send work here from other pages
+
+Most actions come from signals each night, but you can also send work to the Action Center:
+
+- **Mentions**: select answers, or open one, and click **Send to Action Center**.
+- **Content**: **Send to Actions** on a content opportunity, or **Assign to…** on a Content Plan row.
+- **Site Audit**: **Send to Action Center** on a page's audit report.
+- **AI Referral Traffic**: select pages and click **Send to Action Center**.
+- **Brand Visibility**: **Send action** on a topic in Opportunities.
+- **Prompts**: **Send Actions** on a prompt's Improve tab.
+
+Sending the same thing twice doesn't create a second action: the page says it is already in the Action Center, with its action number.
diff --git a/help/action-center/action-center-kpis.md b/help/action-center/action-center-kpis.md
new file mode 100644
index 000000000..4083331ef
--- /dev/null
+++ b/help/action-center/action-center-kpis.md
@@ -0,0 +1,31 @@
+---
+title: Track KPIs in the Action Center
+route: /dashboard/action-center/kpis
+keywords: [kpis, goals, targets, edit kpis, kpi status, goal reached]
+synonyms: [key metrics, objectives, set goals, performance targets]
+plan: all
+updated: 2026-10-29
+last_checked: 2026-10-29
+owner: Cihan
+app_version: 0.3.0
+approved: false
+sources: [docs/guides/action-center.mdx]
+---
+The **KPIs** tab tracks the goals you set for the brand. Each KPI shows its **Value**, **Change**, **Trend (7D)**, **Goal**, progress **vs Goal** and a **Status**. An action's Overview lists the KPIs it is linked to.
+
+## KPIs you can track
+
+AI Visibility, Citations, Mentions, Share of Voice, AI Referral Traffic and Positive Sentiment.
+
+A brand that never had KPIs starts with a suggested set, each aiming 20% above its last 30 days. It is saved the first time an admin or manager opens the tab.
+
+## KPI status
+
+- **Goal Reached**: 100% of the goal or more.
+- **On Track**: 70% or more, and not moving the wrong way.
+- **At Risk**: 40 to 69%, or moving the wrong way before the goal is reached.
+- **Off Track**: under 40%.
+
+## Change your KPIs
+
+Click **Edit KPIs** (or **Configure KPIs**) to open the framework. Pick a template (**AI Search Visibility**, **Brand Monitoring** or **Custom Framework**), a time period (Weekly, Monthly or Quarterly) and a target for each KPI, then click **Save & Apply**. Only admins and managers can change KPIs.
diff --git a/help/action-center/action-center-overview.md b/help/action-center/action-center-overview.md
new file mode 100644
index 000000000..aa2281f71
--- /dev/null
+++ b/help/action-center/action-center-overview.md
@@ -0,0 +1,35 @@
+---
+title: Action Center overview
+route: /dashboard/action-center
+keywords: [action center, actions, signals, kpis, history, prioritized work]
+synonyms: [to-do list, work list, what to do next, recommendations, home page]
+plan: all
+updated: 2026-10-29
+last_checked: 2026-10-29
+owner: Cihan
+app_version: 0.3.0
+approved: false
+sources: [docs/guides/action-center.mdx]
+---
+The **Action Center** is where Ansvisor turns what changed in AI search into a ranked list of work. It is the first page you see after you sign in. Everything in it is per brand: switch brands to see another brand's work.
+
+## The tabs
+
+- **Actions**: what to do now, ranked by Priority Score.
+- **Signals**: what changed across your AI visibility data.
+- **KPIs**: the goals you track for the brand.
+- **History**: what was done, and what it moved.
+
+Admins and managers also see **Needs review**, where tasks the AI wrote without some of their details wait for a person.
+
+## How the work flows
+
+Signals are detected after your daily tracking run. Ansvisor groups related signals into one action, so a long list of findings becomes a few pieces of work. At most five new actions are added per brand per day, so the list stays short enough to act on. Each action comes with tasks. When you close an action as **Completed**, History measures what changed afterwards.
+
+## While the first analysis runs
+
+While a new brand's first analysis runs, a **First analysis running** note sits under the tabs. Actions, signals and KPIs fill in once it finishes. Meanwhile, a new brand starts with the **Getting started** steps on the Actions tab.
+
+## A weekly routine
+
+Work from the **Top Actions** card each week. Assign each action, set a due date, and close it as **Completed** when its tasks are done, so History can measure the result.
diff --git a/help/action-center/action-center-signals.md b/help/action-center/action-center-signals.md
new file mode 100644
index 000000000..0b9868382
--- /dev/null
+++ b/help/action-center/action-center-signals.md
@@ -0,0 +1,45 @@
+---
+title: Signals in the Action Center
+route: /dashboard/action-center/signals
+keywords: [signals, what changed, visibility drop, lost citation, signal status]
+synonyms: [alerts, changes, detections, findings]
+plan: all
+updated: 2026-10-29
+last_checked: 2026-10-29
+owner: Cihan
+app_version: 0.3.0
+approved: false
+sources: [docs/guides/action-center.mdx, docs/guides/ai-traffic-analytics.mdx]
+---
+The **Signals** tab lists what changed: a visibility drop, a lost citation, a competitor gaining, a page with no AI traffic, a site issue. Signals are detected after your daily tracking runs, and they are the evidence actions are built from. Many of the same changes also appear in the Daily Pulse email.
+
+## Cards and filters
+
+The cards show Total, Important (high impact), New and Resolved signals, and the average signal impact, for the selected range (7, 30 or 90 days) against the range before.
+
+Filter by category (Visibility, Citation, Mention, Traffic, Technical, Competitor), **Impact**, **Status**, **Source** (AI Results, Google Analytics, Google Search Console, Site Audit) and **Platform**, or search.
+
+## Signal status
+
+A signal is **New**, **Acknowledged**, **Resolved** or **Dismissed**. Set it from the row menu, and use **Reopen** to bring one back. When you resolve or dismiss a signal, the actions it supports get their Priority Score again right away.
+
+## Inside a signal
+
+Click a signal to see:
+
+- **Where** it was found;
+- **What changed**: before and after, with the periods compared;
+- **Evidence and links**;
+- **Related KPIs**;
+- **Used by actions**: the actions built from it.
+
+Copies of one finding, such as the same prompt text tracked as several prompts, are grouped into one row and counted once.
+
+## AI traffic signals
+
+The nightly pass also watches AI Referral Traffic and raises these signals:
+
+- **AI visits fell from … to … week over week**: AI visits over the last 7 whole days fell 40% or more from the 7 before, which had at least 30. It feeds **Recover AI traffic**.
+- **Broken AI links on …**: AI visitors landed on a missing page in the last 7 days. It feeds **Fix broken citation targets**.
+- **AI search bots haven't read … that AI answers cite**: needs 30 days of AI-bot data. It feeds **Fix AI crawler accessibility**.
+- **… started sending visitors**: an AI platform sent visits for the first time.
diff --git a/help/action-center/action-history-results.md b/help/action-center/action-history-results.md
new file mode 100644
index 000000000..4dacf62b8
--- /dev/null
+++ b/help/action-center/action-history-results.md
@@ -0,0 +1,46 @@
+---
+title: Action history and results
+route: /dashboard/action-center/history
+keywords: [history, action results, scorecard, success rate, measured outcome]
+synonyms: [what worked, past actions, completed actions, did it work]
+plan: all
+updated: 2026-10-29
+last_checked: 2026-10-29
+owner: Cihan
+app_version: 0.3.0
+approved: false
+sources: [docs/guides/action-center.mdx]
+---
+The **History** tab lists actions once they are In Progress, On Hold, Completed or Dismissed, with what they were meant to do and what has been measured since. A result is measured after the action, not proven to be caused by it.
+
+## Find an action in History
+
+Filter by range (7d, 30d, 90d, All), **Status**, **Type**, **Priority** and **Source**, and sort by Newest, Oldest or Priority. The cards count total, completed, in-progress, no-improvement and dismissed actions in the window.
+
+Click an action to see **What was the goal?**, **What happened?** (the signals it was built from), its affected items and tasks, and its **Result**.
+
+## How a result is measured
+
+Only a completed action is measured. Its metrics are read three times, over the week ending 7, 14 and 30 days after it closed, each against the 7 days before it was raised. Until the final reading on day 30, the result reads *so far*, and the drawer lists each reading.
+
+An action that names its prompts or pages is measured on them, with the brand-wide figures beside. Its prompts' change is also shown net of the brand's prompts it didn't touch, so a change that hit every prompt, such as an engine's update, is told apart from its own. A Site Audit fix is judged by its pages audited again since it closed.
+
+## What a result means
+
+- **Improved**, **Mixed**, **No meaningful change** or **Declined**: judged by the action's main metric, the first one it names. **Mixed** means another metric it names moved the other way.
+- **Not measured**: the action was dismissed.
+- **Not measurable**: it has no metric to read, or its main metric is mentions or citations and tracking missed more than two nights in one of its weeks.
+
+The drawer also says what became of the signals that raised it: how many ended, turned into another, were dismissed, or are still open.
+
+## Actions sent from Improve
+
+An action sent from a prompt's Improve tab reports on how many platforms its position improved. It is read again daily while the action is open, and for 30 days after it is completed.
+
+## The Scorecard
+
+Switch History from **Actions** to **Scorecard** to see what each kind of action did at this brand over the range, 90 days unless you pick another. It shows how many were completed, their results so far, the **Success rate** (the share of results that improved, shown once a kind has three), the median change in each action's main metric, and how many were dismissed, with the reason given most often. It counts this brand's own actions only.
+
+## When a problem comes back
+
+When an action is closed, its condition can open a new action after a two-week rest if it is still firing. The old one stays in History as the record of that round of work.
diff --git a/help/action-center/action-priority-score.md b/help/action-center/action-priority-score.md
new file mode 100644
index 000000000..c75f97a01
--- /dev/null
+++ b/help/action-center/action-priority-score.md
@@ -0,0 +1,51 @@
+---
+title: Action Priority Score
+route: /dashboard/action-center/actions
+keywords: [priority score, priority level, critical, why this priority, action ranking]
+synonyms: [action priority, ranking, how actions are sorted, urgency]
+plan: all
+updated: 2026-10-29
+last_checked: 2026-10-29
+owner: Cihan
+app_version: 0.3.0
+approved: false
+sources: [docs/guides/action-center.mdx]
+---
+Every action has a **Priority Score** from 0 to 100: how soon to work on it beside the others. Impact says how much an action could matter; priority weighs impact together with the evidence behind it. The score sets the default order of the **Actions** tab and the **Top Actions** card.
+
+## Priority levels
+
+- **Critical**: 90 and above.
+- **High**: 75 to 89.
+- **Medium**: 50 to 74.
+- **Low**: below 50.
+
+## What goes into the score
+
+- **Impact**: the action's impact grade, high, medium or low.
+- **Demand**: how much its prompts, search queries or pages are searched or visited. Only the most demanded few count. A prompt's measured demand level raises it: High more, Medium a little.
+- **Severity**: a loss ranks above a sharp drop, a drop above a slip, and a slip above a gap or a site issue.
+- **Business value**: how commercial the searches are, or what a page is worth from Google Analytics.
+- **Confidence**: how many independent kinds of finding and data sources agree.
+- **Recoverability**: ground you held and lost is nearer at hand than ground you never held.
+- **Recency**: newer evidence counts more.
+
+Demand or business value that nothing measures neither raises nor lowers the score.
+
+## Past success
+
+On top of these, past success nudges the score: how this brand's own earlier actions of the same kind turned out over the last six months, as the History scorecard counts them. A kind that improved results here sits up to 10 points higher, one that rarely did up to 10 lower. A few results barely move it, and a kind with no results yet doesn't move at all. Other brands' results are never used.
+
+## When the score changes
+
+An action is scored when it is created, and again every night from that night's evidence. When you resolve or dismiss a signal, the actions it supports are scored again right away. An action not scored yet shows a dash and sorts after the scored ones.
+
+An action sent from another page has no detection signals, so it is scored from what it carries, such as the prompts or pages that were sent.
+
+## Why this priority?
+
+The action's **Overview** tab starts with **Why this priority?**: one sentence written from the facts the score rests on. It says what changed, on what, how recently, and what makes it matter. When past success moves the score by a point or more, it says so too.
+
+## Not the Content Plan's score
+
+The Content Plan has its own Priority Score, which ranks the Plan's pages. The two are worked out separately.
diff --git a/help/action-center/action-tasks.md b/help/action-center/action-tasks.md
new file mode 100644
index 000000000..38d32f537
--- /dev/null
+++ b/help/action-center/action-tasks.md
@@ -0,0 +1,78 @@
+---
+title: Action tasks and how they are checked
+route: /dashboard/action-center/actions
+keywords: [tasks, task status, check now, regenerate tasks, done when, how we check]
+synonyms: [steps, subtasks, to-do items, task checks]
+plan: all
+updated: 2026-10-29
+last_checked: 2026-10-29
+owner: Cihan
+app_version: 0.3.0
+approved: false
+sources: [docs/guides/action-center.mdx]
+---
+Each action comes with tasks: concrete steps such as "Refresh content for these prompts" or "Add structured data to these pages". You find them on the action's **Tasks** tab.
+
+## What a task tells you
+
+Open a task to see:
+
+- **What**, **URL** (the exact page), **Where** and **Why**;
+- **Change**: the text before and after, or the text to add;
+- **How**: one step per line;
+- **Done when** and **How we check**;
+- **Type** (Content, On-page, Technical, Outreach, PR or Product), **Effort** and **Role**.
+
+Click **Edit** to change any of them.
+
+## How we check
+
+**How we check** says how a task is confirmed:
+
+- **Automatic**: Ansvisor fetches the page again.
+- **Re-run**: the prompts run again.
+- **Manual**: you confirm it.
+
+Automatic checks run every night, and when you click **Check now**. Only pages on the brand's own site are read. A task that passes is completed by itself, and History says so. One that doesn't shows **Last check** with what is still missing. Site Audit fixes are checked by your next audit.
+
+## What an automatic check looks for
+
+- **A broken page**: the page answers again.
+- **Publishing**: the page answers where it wasn't before.
+- **Structured data**: the audit checks that raised the action now pass on the page.
+- **Crawler access**: robots.txt lets the AI crawlers in, and the page isn't noindex.
+- **Optimizing or refreshing**: the page changed by more than a date, a price or a rotating quote. Blocks that change by themselves, such as the latest posts, don't count.
+- **New text**: the task's new text is on the page, and the text it replaces is gone.
+
+A task you reopen after it passed stays open: the call is yours.
+
+## Task status
+
+A task's status is **To Do**, **In Progress**, **Completed** or **Skipped** (with a reason). Skipped tasks leave the count, so they don't hold an action short of done. Use **Add task** for work of your own, and delete tasks you don't need.
+
+## Hand a task on
+
+From a task you can:
+
+- **Create content** or **Optimize page**: open a draft in the Content Editor for this task. **Open draft** returns to it, and marking the draft ready completes the task.
+- **Send to Content**: send a content task to Content as an opportunity, where a brief can be written for it.
+- **Send to webhook**: post the task to the brand's webhook, for a workflow outside Ansvisor. Set it up under **Content › Webhook Settings** with the **Action Center task sent** event selected.
+
+## Regenerate tasks
+
+Where AI task writing is on for your workspace, **Regenerate tasks** writes the action's tasks again from its evidence, in your brand's terms, by the **Site type** in the brand settings. Done, under-way and hand-added tasks stay. It uses your plan's monthly AI allowance.
+
+The AI also reads this brand's last three closed rounds of the same action. It builds on what improved results here, and doesn't hand out again a task that was done without effect. A task it writes without its URL, Change, Done when or How we check waits on **Needs review**.
+
+## Tasks Ansvisor does itself
+
+Some tasks that only read data, such as checking a page's traffic in your connected Analytics, can be carried out by Ansvisor overnight. The action's History says so. Nothing that would change something outside Ansvisor runs without your approval: such a task waits as **Waiting Approval**.
+
+## Comparing pages overnight
+
+Two tasks compare pages for you overnight:
+
+- **Compare competitor content and coverage** reads up to three pages AI answers cited for the action's prompts, your competitors' first, and your own cited page. It says which checks at least half of their pages pass that yours doesn't.
+- **Find what sets your winning pages apart** compares up to three of your winning pages with your other pages that Site Audit checked. It lists what most winning pages have and the others lack, and the pages missing the most of it.
+
+When there is nothing to compare, the task stays open with the reason, and Ansvisor tries again a week later.
diff --git a/help/action-center/getting-started-steps.md b/help/action-center/getting-started-steps.md
new file mode 100644
index 000000000..7bd9ac09e
--- /dev/null
+++ b/help/action-center/getting-started-steps.md
@@ -0,0 +1,43 @@
+---
+title: Getting started steps in the Action Center
+route: /dashboard/action-center/actions
+keywords: [getting started, setup steps, onboarding checklist, new brand, skip get started]
+synonyms: [first steps, setup list, where do I start, onboarding]
+plan: all
+updated: 2026-10-29
+last_checked: 2026-10-29
+owner: Cihan
+app_version: 0.3.0
+approved: false
+sources: [docs/guides/action-center.mdx]
+---
+A new brand starts with **Getting started**: a fixed list of 17 steps on the Action Center's **Actions** tab. While it runs, the **Getting started** category opens first and lists the steps with their **Stage**, **Status** and **Ready** time, and a **Getting started** card shows how many are done. Click **All** or another category to see your actions.
+
+## The four stages
+
+- **Set up**, ready now: set your KPIs, add at least 3 competitors and 3 topics, add 3 prompts per topic, connect Google Search Console and Google Analytics, and add the AI Traffic snippet.
+- **Explore**, when the first AI answers are in (about 3 hours): see where you stand in Brand Visibility, read your first AI answers, track high-value prompt suggestions, check what AI says about you, see who AI quotes and see visits from AI.
+- **Act**, when the first content suggestions are in (about 24 hours): pick content from suggestions, create your first content, optimize a page.
+- **Ready**, after every step above is done or skipped.
+
+The high-value prompts step waits about 24 hours after you connect Search Console or Analytics.
+
+## Complete a step
+
+Click a step to open its card: why it matters, how to do it, and when it counts as done. The card's button opens the step's page in a new tab. Most steps are checked automatically from your data, such as three competitors saved or a first AI visit detected, so the step updates when you come back to the Action Center tab. A step still waiting for its data says when it opens.
+
+## Skip the Google steps
+
+**Connect Google Search Console** and **Connect Google Analytics** are optional. Admins and managers can **Skip** them. A skipped step counts as complete, and you can still connect later from Settings. Without either one, prompt and content suggestions are less accurate.
+
+## Start or skip Getting started
+
+Admins and managers see **Get Started** or **Skip Get Started** at the top right of the Actions tab.
+
+- **Skip Get Started** puts the actions list back to normal. Steps you finished stay done, and you can start again any time from the same button.
+- **Get Started** appears after a skip, and for an existing brand whose setup looks incomplete: fewer than 3 competitors or topics, no Search Console, or no content in the last 30 days.
+- A brand added to an organization that already has one asks first whether to run Getting started for it. Choose **Run Getting started** or **Skip**.
+
+## Steps are not actions
+
+Getting started steps are not actions: they have no Priority Score and never come from signals. Your actions are listed beside them, under **All** and the other categories.
diff --git a/help/action-center/task-review-queue.md b/help/action-center/task-review-queue.md
new file mode 100644
index 000000000..5d2e0df17
--- /dev/null
+++ b/help/action-center/task-review-queue.md
@@ -0,0 +1,26 @@
+---
+title: Review tasks waiting in Needs review
+route: /dashboard/action-center/review
+keywords: [needs review, approve task, dismiss task, task review, missing fields]
+synonyms: [review queue, pending tasks, tasks to approve, incomplete tasks]
+plan: all
+updated: 2026-10-29
+last_checked: 2026-10-29
+owner: Cihan
+app_version: 0.3.0
+approved: false
+sources: [docs/guides/action-center.mdx]
+---
+A task the AI writes without its **URL**, its **Change**, **Done when** or **How we check** doesn't reach the action's task list. It waits on the Action Center's **Needs review** tab instead, with what it misses. Only admins and managers see the tab and review tasks.
+
+## Approve a task
+
+Click **Edit**, fill the missing fields and click **Approve**. The task is added after the action's other tasks. All four fields, the URL, the change, Done when and How we check, must be filled before it can be approved.
+
+## Dismiss a task
+
+Click **Dismiss** on a task you don't want. A dismissed task never reaches its action.
+
+## Where these tasks come from
+
+They come from AI task writing, such as **Regenerate tasks** on an action. Only the tasks that miss one of the four fields wait here; the others go straight to the action's **Tasks** tab.
diff --git a/help/ai-traffic/ai-bots.md b/help/ai-traffic/ai-bots.md
new file mode 100644
index 000000000..56d060359
--- /dev/null
+++ b/help/ai-traffic/ai-bots.md
@@ -0,0 +1,36 @@
+---
+title: See which AI bots read your site
+route: /dashboard/traffic
+keywords: [ai bots, crawlers, gptbot, claudebot, perplexitybot, cloudflare, vercel]
+synonyms: [bot logs, ai crawlers, who crawls my site, bot traffic]
+plan: growth
+updated: 2026-10-29
+last_checked: 2026-10-29
+owner: Cihan
+app_version: 0.3.0
+approved: false
+sources: [docs/guides/ai-traffic-analytics.mdx]
+---
+The **AI bots reading your site** section of **AI Traffic Analytics** shows which AI bots read your pages, from your own Cloudflare or Vercel. Only verified bots count: the bot's name and its operator's published addresses must both match. Today that's the bots of OpenAI, Anthropic and Perplexity.
+
+## What the section shows
+
+- **By bot type**: hits, pages read and last seen for **Training**, **Search index** and **User fetch**, with the bots under each.
+- **Pages AI bots read**: hits per page, and when each type last read it.
+- **The search-bot alert**: pages AI answers cited in the last 30 days that no AI search bot read in that time. It starts 30 days after the first AI-bot request, and pauses when the bot data doesn't cover the last 30 days or the month's cap was reached.
+
+## Connect AI bots
+
+Click **Connect** on the **AI bots** card in the Data sources strip, and choose how your site sends AI-bot requests: **Cloudflare Worker** (recommended), **Cloudflare Logpush** or **Vercel Drain** (Vercel Pro and Enterprise only). Each option says what it costs on your own Cloudflare or Vercel bill before you connect, and none of them changes your site's speed.
+
+Click **Create connection** and follow the numbered steps. Copy the settings right away: the secret in them isn't shown again. If you lose it, revoke the connection and create a new one.
+
+## Manage connections
+
+**Manage** on the card lists your connections with their last request, lets you **Revoke** one, and shows this month's use. Cloud plans store a monthly number of detailed rows; past it, detailed rows pause until the next month and the totals are still counted. Only requests to your brand's own domains, or their subdomains, are kept.
+
+## Dates and sample data
+
+Bot counts are kept by UTC day, so the 24-hour range reads yesterday and today (UTC), and All time reads the last 90 days.
+
+On the Starter plan, the section shows sample data with **Upgrade**. Without a connection, it offers **Connect** or **Preview with sample data**.
diff --git a/help/ai-traffic/ai-referral-traffic.md b/help/ai-traffic/ai-referral-traffic.md
new file mode 100644
index 000000000..5702ed53e
--- /dev/null
+++ b/help/ai-traffic/ai-referral-traffic.md
@@ -0,0 +1,67 @@
+---
+title: AI Referral Traffic page
+route: /dashboard/traffic
+keywords: [ai referral traffic, ai traffic analytics, ai visits, chatgpt visits, referral trend]
+synonyms: [traffic from ai, visitors from chatgpt, ai referrals, ai sessions]
+plan: all
+updated: 2026-10-29
+last_checked: 2026-10-29
+owner: Cihan
+app_version: 0.3.0
+approved: false
+sources: [docs/guides/ai-traffic-analytics.mdx]
+---
+**AI Referral Traffic** in the sidebar opens the **AI Traffic Analytics** page. It measures real visits to your site from AI assistants, what those visits are worth, and whether AI is changing your Google traffic. Most analytics tools count these visits as direct or referral traffic; Ansvisor tells each assistant apart.
+
+## Date range and changes
+
+Pick a range: 24h, 7d, 30d, 90d, All or Custom. The page opens on the last 24 hours, then on the range you last picked in this browser, and every section follows it.
+
+Each figure is compared with the period of the same length just before. Below 30 visits in either period, the change reads **Low sample** instead of a percentage. A platform or page with no visits before reads **new**. All time has no previous period.
+
+## How visits are counted
+
+- **One visit per session**: a reload, or a further link from the same answer, continues the visit. The visits card gives the pageviews beside the visits.
+- **Snippet or Google Analytics, never both**: on days the tracking snippet sent any visit, the page counts the snippet's; on other days, Google Analytics'. The visits card says how many days came from Google Analytics.
+- **Recent AI Referral Visits**, at the bottom, shows only the snippet's visits, one row per visit. Search it by address and filter by platform.
+
+## Summary line and cards
+
+One or two sentences above the cards sum up the range. Each fact appears only when the data can back it. The cards:
+
+- **AI-referred visits**, with the change on the previous period;
+- **Top platform** and **Platforms**;
+- **AI share of traffic**, **Engaged visits** and **Key events from AI**, which need Google Analytics;
+- **Google clicks**, which needs Search Console. Google reports about two days late, so 24h and 7d show its latest reported days.
+
+## Trend and platforms
+
+- **AI Referral Trend**: **By platform** shows visits per day, or per hour on 24h. **vs Google clicks** sets AI visits beside Google clicks; it needs Search Console and a range of 7 days or longer. Dashed lines mark the days actions entered the Action Center or prompts were added.
+- **Platform Breakdown**: each AI platform's share of visits.
+- **Referrals by Platform**: visits and change per platform, with **Engaged** and **Key events** from Google Analytics.
+- **Top Landing Pages**: the pages with the most AI visits, and how often tracked answers cite them.
+
+AI Overviews and AI Mode clicks are part of Google organic clicks, and Google doesn't let them be separated.
+
+## Where AI visitors convert
+
+With Google Analytics connected, two tables show what AI visits bring:
+
+- **Top converting pages from AI**: landing pages ranked by key events from AI visits, or by revenue when the property reports it, with the best platform.
+- **Which answers bring buyers**: the topics of your tracked prompts whose answers cite those converting pages, by AI platform.
+
+A range with fewer than 30 AI visits uses the last 30 days here, and says so.
+
+## Which AI assistants count
+
+ChatGPT, Perplexity, Claude, Gemini, Copilot, Grok, DeepSeek, Meta AI, You.com, Phind, Poe and Google Labs. A visit counts when it comes from one of their sites, or from a link tagged with their name. Google Analytics' own **AI Assistant** channel also shows assistants that aren't on the list yet. AI visits that arrive with neither count as Direct.
+
+## Export the page
+
+**Export** gives the whole page as **Excel (.xlsx)**, one sheet per section, or **CSV (.zip)**, one file per section. Each table also has its own Excel and CSV buttons. A range with more than 100,000 visits is too large to export at once.
+
+## Weekly AI Traffic e-mail
+
+Every Monday, each brand with AI visits gets an e-mail about the week before, sent to its Daily Pulse recipients. It shows AI visits with the change, the top platform, the summary line, visits by day and platform, and the top changes.
+
+To turn it off for a brand, untick **AI Traffic weekly digest** in **Settings › Notifications**; admins and managers can change it. It isn't sent while the brand's Daily Pulse is **Off**. The e-mail is on Ansvisor Cloud only.
diff --git a/help/ai-traffic/ai-traffic-pages.md b/help/ai-traffic/ai-traffic-pages.md
new file mode 100644
index 000000000..a0867cd3e
--- /dev/null
+++ b/help/ai-traffic/ai-traffic-pages.md
@@ -0,0 +1,46 @@
+---
+title: AI traffic by page and Google query
+route: /dashboard/traffic
+keywords: [no ai traffic, cited not clicked, broken ai links, untracked google queries, pages]
+synonyms: [pages without ai visits, 404 from ai, redirect suggestions, ai taking google clicks]
+plan: all
+updated: 2026-10-29
+last_checked: 2026-10-29
+owner: Cihan
+app_version: 0.3.0
+approved: false
+sources: [docs/guides/ai-traffic-analytics.mdx]
+---
+The **Pages** section of **AI Traffic Analytics** answers which pages get AI visits, which don't, and why. Below it, the Google Search Console tables show whether AI is changing your Google traffic.
+
+## The Pages tabs
+
+- **No AI traffic**: pages with traffic in Google Analytics or in your sitemap that got no AI visit, most valuable first. **AI visibility** says why: **Cited**, **Not cited** or **No prompt**.
+- **Cited, not clicked**: pages AI answers cited at least 20 times in the range that got no AI visit.
+- **Traffic, no prompt**: pages AI assistants sent visitors to that no tracked prompt's answer cited.
+- **Broken AI links**: pages AI visitors landed on that don't exist.
+- **All pages**: every page with AI visits, citations, traffic or a sitemap entry in the range.
+
+## Next steps from a page
+
+Rows carry their next step: **Audit page** for a page AI cites or a prompt targets, **Add prompts** for one no prompt points to.
+
+On **No AI traffic** and **Cited, not clicked**, each row's **…** menu also has **Send to Action Center**, **Add prompts** and **Suggest content**. Under the list, **Audit N cited pages** (or **Audit N pages**) starts Site Audits for the pages shown, and **Send to Action Center** sends them as one action.
+
+## Add prompts for a page
+
+**Add prompts** suggests questions people ask AI assistants that the page answers, including ones from its Search Console queries, and lets you add your own. The prompts you track from it point to the page, so their answers show whether AI cites it. Each one uses a prompt slot on your plan.
+
+## Fix broken AI links
+
+**Broken AI links** shows what the visitor got: **Not found (404)**, **Gone (410)**, **Shows a not-found page (200)** for a site that shows its own not-found page, or **Page title says not found** where the browser reports no status.
+
+Each comes with a suggested **Redirect to**: a **Similar page**, **The page above it** or the **Home page**. Addresses with a query string get no suggestion. With the tab open, the Excel and CSV buttons export the redirect list. Add the tracking snippet to your 404 page too, so these visits are seen.
+
+## Is AI taking Google clicks?
+
+With Search Console connected, **Is AI taking Google clicks?** lists pages whose impressions rose 15% or more while their click-through rate fell 20% or more, among pages with at least 100 impressions in the period before. An AI Overview may be answering in place of the click. **Positions** shows whether you're cited for a tracked prompt about the page; **Add prompts** appears when none is about it yet.
+
+## Untracked Google queries
+
+**Untracked Google queries** lists Search Console queries with clicks that no tracked prompt covers, most clicked first. **Track** adds one as a prompt; **Track all** adds the listed ones that your plan's prompt slots allow. Admins and managers can track queries.
diff --git a/help/ai-traffic/ai-visit-conversions.md b/help/ai-traffic/ai-visit-conversions.md
new file mode 100644
index 000000000..0dce354d5
--- /dev/null
+++ b/help/ai-traffic/ai-visit-conversions.md
@@ -0,0 +1,54 @@
+---
+title: Track conversions from AI visits
+route: /dashboard/traffic
+keywords: [conversions, ai visit conversions, signups from ai, consent, shopify pixel]
+synonyms: [track leads from chatgpt, orders from ai, demo requests from ai, first touch]
+plan: all
+updated: 2026-10-29
+last_checked: 2026-10-29
+owner: Cihan
+app_version: 0.3.0
+approved: false
+sources: [docs/guides/ai-traffic-analytics.mdx]
+---
+Google Analytics' key events say how many AI sessions converted. To see which pages AI visitors went through on the way to a signup, a demo form or an order, turn on conversions for the brand. The snippet then follows visitors who came from an AI assistant for 30 days, and credits each conversion to the assistant that first sent them.
+
+## Turn conversions on
+
+In **Brand settings › Tracking**, under **Conversions from AI visits**, tick **Track conversions from AI visits** and choose **Consent**:
+
+- **Wait for the visitor's consent** (recommended): no journey is followed until your site says the visitor agreed. AI visits themselves are counted as before.
+- **Don't wait: our own banner covers it**: only if your cookie banner already asks for analytics storage, or where no consent is needed.
+
+Conversions are off by default, and only admins and managers can change them.
+
+## Replace the tag
+
+Under **Use this tag**, copy the tag with the copy button on its code, and use it in place of the one you have, on every page. The tag carries your choices, so after you change **Consent**, copy it again.
+
+To turn conversions off, untick **Track conversions from AI visits** and put the tag without them back.
+
+## Tell Ansvisor when a visitor agrees
+
+With **Wait for the visitor's consent**, your cookie banner has to tell the snippet when the visitor agrees, and when they say no or take consent back. Google Consent Mode needs nothing more. A banner set up inside Google Tag Manager needs a tag there for agreeing and one for refusing.
+
+The visitor has to agree on the page the AI assistant sent them to: agreeing on a later page doesn't bring the journey back.
+
+## What counts as a conversion
+
+- A form with an email field counts as a signup on its own, once per page load. Contact and newsletter forms count too, unless you give them their own goal. Login forms don't count, and nothing typed in a form is sent.
+- Anything else, such as a thank-you page or a purchase, needs a conversion call from your site's own code.
+- On Shopify, orders come from a Customer events pixel. Copy it under **Shopify store** in **Brand settings › Tracking** and add it in Shopify admin as a custom pixel.
+
+## Where you see the results
+
+The Content Editor shows what came of it: the **Live page** tab for each page, and the Content plan's AI results line for the whole site.
+
+## What it can't see
+
+- Visitors who refuse cookies, or who agree only after leaving the page the AI assistant sent them to.
+- AI visits that arrive without a referrer: they count as Direct.
+- Google's AI Overviews and AI Mode.
+- Journeys from one subdomain to another.
+- Forms inside an embedded widget, forms sent from a script, and posts without a form. These need a conversion call from your site's code.
+- Conversions after Safari has cleared the visitor's id, which it does after seven days of use without a visit to your site.
diff --git a/help/ai-traffic/connect-ai-traffic-sources.md b/help/ai-traffic/connect-ai-traffic-sources.md
new file mode 100644
index 000000000..d2e787292
--- /dev/null
+++ b/help/ai-traffic/connect-ai-traffic-sources.md
@@ -0,0 +1,47 @@
+---
+title: Connect AI traffic data sources
+route: /dashboard/traffic
+keywords: [tracking snippet, connect google analytics, connect search console, data sources, install snippet]
+synonyms: [set up ai traffic, add tracking code, link analytics, ga4, gsc]
+plan: all
+updated: 2026-10-29
+last_checked: 2026-10-29
+owner: Cihan
+app_version: 0.3.0
+approved: false
+sources: [docs/guides/ai-traffic-analytics.mdx]
+---
+The **Data sources** strip at the top of **AI Traffic Analytics** shows the four sources the page reads, and each one's state. A section whose source isn't connected stays on the page with a **Needs …** tag and a **Connect** button. Only admins and managers can connect a source.
+
+## What each source adds
+
+- **Tracking snippet**: AI-referred visits, page by page, as they happen.
+- **Google Analytics**: AI visits on days the snippet has none, plus engaged visits, key events, revenue and the AI share of all traffic.
+- **Google Search Console**: Google clicks, the pages where AI may be taking them, and the Google queries you don't track yet.
+- **AI bots**: which AI bots read your pages, from your own Cloudflare or Vercel, on the Growth plan and above.
+
+## Add the tracking snippet
+
+Copy the snippet from the page. It sits at the top until it has sent a visit in the last week; after that, **Code** in the Data sources strip opens it. The page fills in your own address and tracking code, so copy it rather than typing it.
+
+Add it to the head of every page you want to track, your 404 page included, so **Broken AI links** can see visitors who land on a missing page.
+
+## Check that the snippet works
+
+The **Tracking snippet** card reads **Active · last visit …** once the snippet has sent a visit in the last week, otherwise **Not detected**. A visit is stored as soon as an AI-referred visitor lands. Reload the page, or change the range, to see new ones.
+
+## What the snippet collects
+
+The snippet sends nothing for a visitor who didn't come from an AI assistant. For an AI-referred visit it sends the page address, the referrer, the AI source, the browser, its language and screen size, and the page's status as the browser saw it. Ansvisor adds the country.
+
+The snippet sets no cookies, and without conversions it stores nothing in the visitor's browser. The visitor's IP address is never stored.
+
+## Connect Google Analytics and Search Console
+
+Click **Connect** on either card in the Data sources strip and sign in to Google; access is read-only. If the card then reads **Connected · choose a property for this brand**, pick the property under **Manage**, in **Settings › Integrations**. Connections are shared across your organization.
+
+On a self-hosted server, both cards read **Not set up on this server** until the server has its Google connection keys.
+
+## When the first numbers arrive
+
+Google Analytics' first numbers arrive in about 15 minutes; up to 14 months of history can take up to 24 hours. Search Console's can take up to 48 hours, because Google reports about two days late.
diff --git a/help/citations/citations.md b/help/citations/citations.md
new file mode 100644
index 000000000..8fa5c2262
--- /dev/null
+++ b/help/citations/citations.md
@@ -0,0 +1,32 @@
+---
+title: Citations page
+route: /dashboard/citations
+keywords: [citations, cited sources, cited URLs, source types, domains]
+synonyms: [which sites AI cites, AI sources, links in AI answers, references in AI answers]
+plan: all
+updated: 2026-10-29
+last_checked: 2026-10-29
+owner: Cihan
+app_version: 0.3.0
+approved: false
+sources: [docs/guides/citations.mdx]
+---
+The **Citations** page gathers the URLs that AI engines cited across your tracked prompts, with its domain and the number of times it was cited. It shows which sources AI engines rely on when they answer questions in your category, and sorts them by source type.
+
+## Domains and URLs
+
+Under **Top Sources**, the **Domains** tab lists the cited sources by domain and the **URLs** tab by single page. Four buttons above them choose the sources listed: **Brand** (your own domains, the default), **Competitors**, **Third-Party** or **All**. Click a URL to open its own page, which shows the AI answers that cited it and the prompt behind each. **Source Types** shows how the cited sources split by type.
+
+## Source types
+
+Ansvisor classifies every cited source automatically. Your own pages and your competitors' pages are marked as such, and other sources get a type such as **Review**, **Forum** or **Social**. The types matter differently:
+
+- Your own pages are the easiest to optimize.
+- Review sites are weighted heavily by AI engines for buying-intent prompts.
+- Forums are surprisingly influential for technical and B2B questions.
+- Social sites carry the lowest weight in current AI models.
+- When AI cites competitors' pages, you are losing ground in comparisons.
+
+## Find competitor pages AI cites
+
+Above the Domains and URLs lists, choose **Competitors** to see the competitor pages AI engines cite when they answer questions in your category. Each one is an opportunity for a counter-piece: content of your own that answers the same question.
diff --git a/help/citations/cited-url-details.md b/help/citations/cited-url-details.md
new file mode 100644
index 000000000..90eab03fc
--- /dev/null
+++ b/help/citations/cited-url-details.md
@@ -0,0 +1,22 @@
+---
+title: Details of a cited URL
+route: /dashboard/citations/url
+keywords: [cited URL, URL details, citation drill-down, cited in, outreach]
+synonyms: [which answers cite this page, why is this page cited, prompts citing a URL, citation details]
+plan: all
+updated: 2026-10-29
+last_checked: 2026-10-29
+owner: Cihan
+app_version: 0.3.0
+approved: false
+sources: [docs/guides/citations.mdx]
+---
+Every URL that AI engines cite has its own page in Ansvisor. Open it from the **URLs** tab of the Citations page by clicking a URL. Its header shows the page's title, its address and its source type.
+
+## Answers that cited it
+
+**Cited in** lists every AI answer that cited the URL in the selected window, with the date, the AI platform and the tracked prompt that triggered the citation. Use it to see exactly which questions lead AI engines to this page.
+
+## Turn it into outreach
+
+A review or list article that AI engines cite but that doesn't name your brand is a direct outreach target: ask its publisher to include you. If the URL is a competitor's page, it is an opportunity for a counter-piece of your own that answers the same question.
diff --git a/help/competitors/add-competitors.md b/help/competitors/add-competitors.md
new file mode 100644
index 000000000..12b2bd9ae
--- /dev/null
+++ b/help/competitors/add-competitors.md
@@ -0,0 +1,26 @@
+---
+title: Add competitors to track
+route: /dashboard/competitors
+keywords: [add competitor, track competitors, competitor name, competitor domain]
+synonyms: [new competitor, follow a rival, competitor list, which competitors to track]
+plan: all
+updated: 2026-10-29
+last_checked: 2026-10-29
+owner: Cihan
+app_version: 0.3.0
+approved: false
+sources: [docs/guides/competitors.mdx]
+---
+Competitors are the brands Ansvisor compares with yours in AI answers to your tracked prompts. You add them on the **Competitors** page, and each one then appears in **Tracked Competitors** for a head-to-head comparison.
+
+## Add a competitor
+
+Click **Add Competitor**. Enter the **Name** exactly as it appears in AI answers: the brand name, not its web address. The **Domain** is optional and is used to cross-check citations. Click **Add**. When a competitor is added, Ansvisor reads back 90 days of your answers once, to find the first time AI cited one of its pages.
+
+## How many to track
+
+Aim for three to eight competitors. With fewer than three, the benchmarks have little signal; with more than eight, the comparison view gets noisy.
+
+## Find competitors you missed
+
+Even if you don't add them, AI answers to your prompts name other brands. Review those names once a month: a brand you've never heard of that is cited three or more times across your prompts is a competitor AI engines already know about. Add it to your list.
diff --git a/help/competitors/compare-competitors.md b/help/competitors/compare-competitors.md
new file mode 100644
index 000000000..64a28d0e1
--- /dev/null
+++ b/help/competitors/compare-competitors.md
@@ -0,0 +1,40 @@
+---
+title: Compare your AI visibility with competitors
+route: /dashboard/competitors
+keywords: [competitors, head-to-head, competitor comparison, gaps, strengths, platform breakdown]
+synonyms: [benchmark competitors, how do we compare, competitor visibility, us versus them]
+plan: all
+updated: 2026-10-29
+last_checked: 2026-10-29
+owner: Cihan
+app_version: 0.3.0
+approved: false
+sources: [docs/guides/competitors.mdx]
+---
+The **Competitors** page compares the AI visibility of the competitors you track with yours, across your tracked prompts and AI platforms. **Tracked Competitors** lists them with their visibility; select one to see it head-to-head with your brand.
+
+## Head-to-head comparison
+
+The comparison shows **Your Avg. Score**, **Their Avg. Score** and the **Score Difference**. **Gaps** lists the prompts where the competitor leads you, and **Strengths** the prompts where you lead. In **Prompt-by-Prompt Comparison**, click a prompt to expand it and see your visibility against the competitor on that single prompt, platform by platform.
+
+## Platform Breakdown
+
+Some competitors dominate one AI platform but underperform on another. **Platform Breakdown** shows these differences, so you can aim your work at the platforms where a competitor is ahead.
+
+## The competitor's pages over time
+
+Below the comparison, a card shows the competitor's pages over time: why it may be rising. Its chart shows the answers that mentioned the competitor, and your brand, each week for the last twelve weeks, with a marker on each week something happened to its pages. The events are listed newest first:
+
+- **First cited in your tracked answers**: the first day an AI answer to one of your prompts cited one of its pages.
+- **Changed**: once a week, Ansvisor reads again the 10 competitor pages your answers cited most over the last two weeks. When a page's words changed by more than a date or a price, it says so, with the Site Audit checks the page gained or lost, such as an FAQ or a table.
+
+## How the pages are read
+
+Pages are read for free from their public address, within a few reads a day, never from a site whose robots.txt keeps crawlers out. The brand's daily page reads are shared with the Action Center's tasks, which go first. Only what Site Audit's checks found and a summary of the page's words are kept, never its text. Nothing is compared with another brand's answers.
+
+## Close a visibility gap
+
+1. Find the prompts where the competitor leads most, under **Gaps**.
+2. Inspect the AI answer: which content or citation is the competitor winning with?
+3. Create content that targets that prompt.
+4. Publish it, earn placements, and watch the gap for two to four weeks to see it close.
diff --git a/help/content/connect-a-cms.md b/help/content/connect-a-cms.md
new file mode 100644
index 000000000..6c10aa1ff
--- /dev/null
+++ b/help/content/connect-a-cms.md
@@ -0,0 +1,54 @@
+---
+title: Connect Webflow, WordPress, Framer or Wix
+route: /dashboard/settings/integrations
+keywords: [connect cms, webflow, wordpress, framer, wix, integrations, application password]
+synonyms: [cms integration, link my blog, set up send to cms, connect my website]
+plan: all
+updated: 2026-10-29
+last_checked: 2026-10-29
+owner: Cihan
+app_version: 0.3.0
+approved: false
+sources: [docs/guides/send-to-cms.mdx]
+---
+To send Content Editor pages straight to your CMS, connect it in **Settings › Integrations** with the brand selected. Each brand connects its own Webflow site, WordPress site, Framer project or Wix site. Only admins and managers can connect, change or disconnect a CMS; members see where the brand's pages go.
+
+## Connect Webflow
+
+In the **Webflow** card, open **How to get a site token** and follow its steps to generate a site token in Webflow. Webflow shows the token only once. Paste it in **Site token** and click **Connect**.
+
+Then pick the **Collection** your pages go to, your blog's for example. Under **Which field takes what**, choose the field for the page content and, if you like, the meta title, meta description, structured data and date. The page's title and slug always go to the item's Name and Slug. Click **Save**. If the collection requires a field Ansvisor doesn't fill, **Save** names it: make it optional in Webflow first.
+
+## Connect WordPress
+
+In WordPress, sign in as the user pages should be written by. A separate user with the **Author** role is enough, since Ansvisor writes drafts only. Under **Users › Profile**, add an **Application Password** named Ansvisor and copy it; WordPress shows it once.
+
+In the **WordPress** card, enter the **Site address** as you visit the site, the **User** and the **Application Password**, and click **Connect**. Paste the Application Password, not the user's own password. The site must use https. Ansvisor checks the site before it keeps anything.
+
+## If WordPress won't connect
+
+A message says why. The most common fixes:
+
+- WordPress refused the user or password: check the user name, and paste the Application Password.
+- There's no WordPress at that address: check the address, and in WordPress set **Settings › Permalinks** to anything but **Plain**.
+- Application Passwords are turned off: update WordPress, and allow them in your security plugin.
+- The site refused the request: a firewall or security plugin is blocking it, so allow it.
+- This user can't write posts: use an Author, an Editor or an Administrator.
+
+## Connect Framer
+
+In Framer, open the project's **Site Settings › General** and make a key, as a user who can edit the CMS, then copy it. In the **Framer** card, paste the **Project address** as your browser shows it in Framer, and the key, and click **Connect**.
+
+Pick the **Collection** your pages go to, then which field takes the title and the page content and, if you like, the meta title, meta description and structured data. Click **Save**. A collection with a required field Ansvisor can't fill is marked in the list and can't be saved: make that field optional in Framer first.
+
+## Connect Wix
+
+The site needs the Wix Blog: pages go there as draft posts. In Wix, the site's owner opens **Account Settings › API Keys Manager**, clicks **Generate API Key**, picks the site, gives the key **Manage Blog** and **Manage Ricos Document**, and copies it; Wix shows it once. In the **Wix** card, paste the **Site id** (or the whole address of the site's dashboard) and the **API key**, and click **Connect**. A Blog with no post can't be connected yet: write one post on it as the site's owner first.
+
+## Replace or disconnect
+
+Each card has its own way to change the sign-in: **Replace token** for Webflow, **Change sign-in** for WordPress, **Replace key** for Framer and **Change API key** for Wix. **Disconnect** removes the connection; pages already sent stay in your CMS. Framer and Wix don't let Ansvisor revoke a key, so delete it there too: in the project's **Site Settings › General**, or in Wix's **API Keys Manager**. If the CMS refuses the saved sign-in, the card says so, for example **Sign-in refused** or **Key refused**, and asks for a new one.
+
+## Send through your workflow instead
+
+Pages can also go to your workflow tool, such as n8n, Make or Zapier, which creates the post in your CMS and decides whether it lands as a draft or goes live. On the **Opportunities** page, open **Webhook Settings**, enter your workflow's address in **Webhook URL**, tick **Content page sent**, which is off by default, and save. Pages you send to **Your workflow** then reach it.
diff --git a/help/content/content-briefs.md b/help/content/content-briefs.md
new file mode 100644
index 000000000..3b877018b
--- /dev/null
+++ b/help/content/content-briefs.md
@@ -0,0 +1,40 @@
+---
+title: Generate a content brief
+route: /dashboard/content
+keywords: [content brief, generate brief, outline, target keywords, send to workflow]
+synonyms: [writing brief, article brief, content outline, brief for writers]
+plan: all
+updated: 2026-10-29
+last_checked: 2026-10-29
+owner: Cihan
+app_version: 0.3.0
+approved: false
+sources: [docs/guides/content-optimization.mdx]
+---
+A content brief tells a writer what to write to close an opportunity's gap: the title, the type of content, the outline and the keywords. Ansvisor writes it with AI from the opportunity's data, and you can send it on to the tool your team works in.
+
+## Generate a brief
+
+On **Opportunities**, open an opportunity. You see its full source data and any brief it already has. If it has none, an admin or a manager clicks **Generate Brief**. Once it's written, the button reads **Brief Generated**.
+
+## What a brief holds
+
+- a suggested article title;
+- the content type, such as a blog post, a guide, a comparison page or an FAQ page;
+- a target word count;
+- an outline of three to eight sections, each with its key points;
+- three to ten target keywords, for search and AI answers alike;
+- competitor insights: what rivals do well, where the gap is and what to attack;
+- a recommended call to action.
+
+## Send it to your workflow
+
+Open the opportunity's **…** menu and click **Send to Workflow**. Ansvisor sends the opportunity and its brief to the workflow tool set up in **Webhook Settings**, and its status changes to **Sent to Workflow**. To send several at once, tick them on the Opportunities list and click **Send All**.
+
+## Set up your workflow once
+
+Click **Webhook Settings** in the toolbar of the Opportunities page. Enter the address of your workflow tool's automation, such as an n8n, Make or Zapier flow, in **Webhook URL**. Turn it **Active** and click **Save Webhook**. Only admins and managers can change these settings. From then on, you can send briefs from any opportunity.
+
+## After you publish
+
+When your team has published the article, mark the opportunity done: tick it on the Opportunities list and click **Done** (admins and managers).
diff --git a/help/content/content-editor.md b/help/content/content-editor.md
new file mode 100644
index 000000000..9ede201b3
--- /dev/null
+++ b/help/content/content-editor.md
@@ -0,0 +1,63 @@
+---
+title: Write a page in the Content Editor
+route: /dashboard/content/editor
+keywords: [content editor, draft, placeholders, mark ready, mark published, history, export]
+synonyms: [write an article, edit a draft, page editor, finish a page]
+plan: all
+updated: 2026-10-29
+last_checked: 2026-10-29
+owner: Cihan
+app_version: 0.3.0
+approved: false
+sources: [docs/guides/content-editor.mdx]
+---
+The Content Editor is where pages are written. It opens from the Library, from the Plan, or when you start a page in New content. **Editor** in the sidebar takes you back to the draft you last opened for this brand in this browser. The page saves as you write.
+
+## The header
+
+The header shows the title, the page type, a **Rules** chip with how many rules the page meets now, and whether your changes are saved. Its buttons are **History**, **Copy HTML**, **Export**, **Send to CMS** (admins and managers) and **Mark ready**, and a button to show or hide the side panel.
+
+## Write and paste
+
+The page starts with its H1, which is also its title. A page has one H1: the toolbar won't add a second one. A paste keeps headings, lists, links and tables, and drops fonts, colors and styles. Press Command or Ctrl + Shift + V to paste plain text.
+
+**Attach a file** inserts Excel, CSV, Word, Markdown, HTML or text, up to 10 MB each; a sheet comes in as a table. **Add HTML** adds an embed, a widget or a form, kept exactly as written; Ansvisor never runs it. **Edit as HTML** shows the page's source; scripts, iframes and inline styles are removed when you switch back.
+
+## Placeholders in brackets
+
+Words in square brackets, such as [figure, source] or [competitor], are placeholders: the Content Agent leaves one where a real name, quote or number goes. Each is tinted amber, and the toolbar counts those left to fill. Click the count, or the arrows beside it, to go to the next or previous one, already selected, so you can type over it.
+
+A placeholder in an image's alt text, the structured data or an HTML block is counted too, and **Show in HTML** selects it in the HTML view. Send to CMS waits until none is left.
+
+## Brief and Sources
+
+The side panel's tabs are **Instructions**, **Sources** and **Live page**. **Instructions** opens on **Rules**, with **Brief** beside it.
+
+- **Brief**: the page type, target URL, goal, audience, length, outline and keywords, which **Edit brief** changes. **Rebuild brief** writes a new outline from the sources you keep, each H2 answering a search AI ran for your prompts. **Rewrite from brief**, for a new page, or **Suggest changes from brief**, for an existing one, hands the page to the Content Agent; the page you have now stays in History.
+- **Sources**: the prompts, fan-outs, cited pages and Search Console queries the draft is built from. Uncheck one to leave it out, or **Add source**, then **Rebuild with selection**.
+
+## Live page
+
+For a draft that optimizes a live page, **Live page** shows how that page does now: **Citations**, **AI visits**, **Google clicks** and **Prompts cited**, over 7 days to a year, against the period before. It is not a forecast of what the draft will get. A new page has nothing to measure yet.
+
+Prompts that stopped citing the page are marked **Lost**. **Use the lost prompts in this draft** adds them to the sources and rebuilds the brief. With conversions from AI visits on, **Conversions from AI journeys** shows what the journeys through the page brought in.
+
+## Passages AI cites are locked
+
+When a draft optimizes a live page brought in by its URL, the passages AI answers quoted in the last 30 days are marked in green, and a chip in the header counts them. The Content Agent never changes these passages. If you edit one, the editor asks first: **Keep wording** or **Edit anyway**. Changing a quoted passage may cost the page those citations.
+
+## History, Copy HTML and Export
+
+- **History** keeps earlier versions for 90 days: the live page as imported, the page before agent changes, each Mark ready and publish, and the page before a restore. **Restore** makes a new version, so nothing is lost.
+- **Copy HTML** copies the whole page as HTML, HTML blocks and schema included, to paste into your CMS.
+- **Export** downloads the page as **Word**, **Excel**, **CSV**, **HTML** or **Markdown**. Only the HTML file carries the schema.
+
+## Add products from your catalog
+
+On an online shop's draft, **Products** in the toolbar adds products from your catalog: search by name, pick one to four that are in stock, and click **Insert**. One shows as a card, more as a grid, each with its picture, name and price, linked to its page. A card keeps its price: to show today's, delete the card and add the product again.
+
+## Mark ready and Mark published
+
+**Mark ready** saves the page, sets it to **Ready** and keeps a version in History. If the draft was sent to the Action Center, it also completes its task there. Ansvisor doesn't publish the page: copy or export it to your site, or send it to your CMS.
+
+Once the page is live, **Mark published**, in the Library's **Status** menu or under **Next step** in the header, asks for its live address, and **Live page** measures that address from then. A page your CMS reports live gets its address on its own. Click **Ready** in the header to move the page back to **In review**.
diff --git a/help/content/content-library.md b/help/content/content-library.md
new file mode 100644
index 000000000..df6f8a399
--- /dev/null
+++ b/help/content/content-library.md
@@ -0,0 +1,52 @@
+---
+title: Manage drafts in the Content Library
+route: /dashboard/content/library
+keywords: [content library, drafts, draft status, archive, export drafts]
+synonyms: [content list, my articles, saved pages, draft list]
+plan: all
+updated: 2026-10-29
+last_checked: 2026-10-29
+owner: Cihan
+app_version: 0.3.0
+approved: false
+sources: [docs/guides/content-editor.mdx]
+---
+The **Library**, under **Content** in the sidebar, lists every page you, your teammates and the Content Agent wrote or changed in Ansvisor. Click a row to open the draft in the editor.
+
+## The tabs
+
+Each tab shows its count:
+
+- **All**: every draft that isn't archived.
+- **In progress**: drafts being written.
+- **In review**: drafts the Content Agent wrote, waiting for a person.
+- **Ready**: drafts marked ready.
+- **Published**: drafts marked published, by you or your CMS.
+- **Archived**: drafts sent to the archive. They keep their status for when they're restored.
+
+## Columns and filters
+
+The columns are Content, Type, Method (Create or Optimize), By (you, the Content Agent or a teammate), Status, Created and Updated. The **Created** filter narrows the list to the last 7, 30 or 90 days. Once a page has been sent to your CMS, the **Last send** filter lists the pages by how their last send went: never sent, sending, sent, failed or published.
+
+## The row menu
+
+- **Status**: set the draft to In progress, In review, Ready or Published. Ready works like the editor's Mark ready, and Published asks for the page's live address. Moving a ready or published draft back to In progress or In review opens its task again.
+- **Export Word**, **Export Excel**, **Export CSV**, **Export HTML** or **Export Markdown** for that draft.
+- **Send to CMS**: send the draft to your CMS. Tick several rows to send them at once.
+- **Send to Archive** or **Restore**.
+- **Delete**: the draft leaves the Library for good, with its versions and suggestions. An Action Center task it was sent as stays.
+
+Send to CMS, archiving and deleting are for admins and managers.
+
+## Export the list
+
+**Export** in the header downloads the drafts listed, that is the current tab within the **Created** range and the **Last send** filter:
+
+- **Export CSV** and **Export Excel**: one row per draft, with its title, status, method, page type, author, dates, page address, length and text.
+- **Export Word**: one document, each draft starting on a page of its own.
+
+## Shopping path check for online shops
+
+For an online shop, the **Shopping path** column shows what the last check found for each draft with a shopping path. After each catalog sync, every such draft is read again. A draft needs attention when it links a product that is sold out, refused by Merchant Center or gone from your catalog; hover the mark to see the products. The **Shopping path** filter lists the drafts that need attention, are fine, or have none.
+
+When a published page links a product that no longer sells, a **Fix** action opens in the Action Center, and **Swap in the editor** opens the draft.
diff --git a/help/content/content-opportunities.md b/help/content/content-opportunities.md
new file mode 100644
index 000000000..2a8a85edd
--- /dev/null
+++ b/help/content/content-opportunities.md
@@ -0,0 +1,39 @@
+---
+title: Find content opportunities
+route: /dashboard/content
+keywords: [content opportunities, opportunity score, impact, visibility gap, content ideas]
+synonyms: [content optimization, content gaps, what to write, content suggestions]
+plan: all
+updated: 2026-10-29
+last_checked: 2026-10-29
+owner: Cihan
+app_version: 0.3.0
+approved: false
+sources: [docs/guides/content-optimization.mdx, docs/guides/content-editor.mdx]
+---
+The **Opportunities** page, under **Content** in the sidebar, lists content opportunities: prompts where competitors are more visible than you in AI answers. Each one suggests a piece of content that could close the gap, and you can start writing it, plan it or send it on from there.
+
+## What an opportunity shows
+
+Each opportunity has a title, the angle of the content to write, and an **Impact** of **High**, **Medium** or **Low**, based on the prompt's volume and the size of your visibility gap. It also has an opportunity score from 0 to 100.
+
+Open an opportunity to see its **Source Data**: the **Related Prompt**, **Est. AI Volume**, **Current Visibility**, **Competitor Gap**, **Search Intent**, **Keywords** and **Competitors Cited**.
+
+## Act on an opportunity
+
+- **Send to Editor** starts a page from the opportunity in the Content Editor.
+- **Send to Plan** (admins and managers) puts it at the top of the Content plan's priority list.
+- **Generate Brief** (admins and managers) writes a content brief for it: a title, an outline, keywords and more.
+- **Send to Workflow**, in the opportunity's **…** menu, sends it and its brief to your workflow tool. Its status becomes **Sent to Workflow**.
+
+## Work on many at once
+
+On the list, tick several opportunities. **Send All** sends them all to your workflow, which suits kicking off a content sprint. Admins and managers also see **Dismiss All**, which sets a batch aside when it isn't relevant this quarter, and **Done**, which marks the ones your team has published.
+
+## Owned and earned
+
+Each opportunity has a **Type**. **Owned** means content on your own site. **Earned** means coverage on other sites, such as articles, reviews and links. Some gaps are best closed with a guest post or a press placement rather than a new page of your own.
+
+## Pick what to work on
+
+Don't ship every brief. Editorial time is usually the bottleneck, so pick the two to four high-impact opportunities a week your team can really write, and leave the rest for later.
diff --git a/help/content/content-plan.md b/help/content/content-plan.md
new file mode 100644
index 000000000..6dd14d3d9
--- /dev/null
+++ b/help/content/content-plan.md
@@ -0,0 +1,65 @@
+---
+title: Use the Content plan
+route: /dashboard/content/plan
+keywords: [content plan, priority list, priority score, create optimize protect, losing ground]
+synonyms: [what to write next, content roadmap, content priorities, editorial plan]
+plan: all
+updated: 2026-10-29
+last_checked: 2026-10-29
+owner: Cihan
+app_version: 0.3.0
+approved: false
+sources: [docs/guides/content-editor.mdx]
+---
+The **Content plan**, **Plan** under **Content** in the sidebar, ranks what to create, optimize or protect next. It covers the last 7, 30 or 90 days: pick the period at the top right.
+
+## The four cards
+
+- **Create**: new pages AI needs, that is open opportunities with no page on your site yet.
+- **Optimize**: pages losing ground, and opportunities to improve a page you already have.
+- **Protect**: your top pages, cited at least five times in the period and not losing ground, up to five, and opportunities to defend a page.
+- **Priority**: rows your team marked as a priority.
+
+Click a card to filter the list to it, and click it again to clear the filter.
+
+## When a page is losing ground
+
+A page is losing ground when it is cited at least 20% less per tracked answer than in the period before, from at least three citations. Because the comparison is per tracked answer, a week with fewer runs, or prompts added or removed, doesn't move a page. If tracking began during the earlier period, a note above the list says so.
+
+## The priority list
+
+Each row shows the **Priority Score**; the **Page**, with its title, page type and path, or **New page**; the **Action**: Create, Optimize or Protect; **Why** it's listed, tagged with where that comes from (**Opportunity**, **Citations** or **Protect**); and **Who** is on it: you, the Agent, a teammate or the person it's assigned to.
+
+The row's button, **Create**, **Optimize** or **View**, opens New content on that row, and **Open draft** opens the draft already being written. Rows marked as a priority come first, the rest by score. The list shows up to 50 rows.
+
+## How the Priority Score works
+
+The score runs from 0 to 100 and is built from four factors:
+
+- **Severity**: how urgent the row is. A page losing citations counts most, then a new page AI needs, a page to improve and a page to protect.
+- **Opportunity**: the score of the content opportunity behind the row.
+- **Page value**: the page's AI visits and, with conversions on, the conversions from AI journeys through it.
+- **Citations**: how often AI answers cited the page.
+
+A row without an opportunity, a new page, or a brand with no AI traffic data counts as neutral, not zero. Hover the score to see **How this score is worked out**. Only a page losing citations can reach critical.
+
+## The row menu
+
+- **Assign to…**: for a row with a draft, pick a teammate and the draft goes to the Action Center as their task. For a row without one, New content opens with **Team** chosen.
+- **Mark as priority** or **Remove priority**: the row moves to the top of the list, tagged **Priority**. **Send to Plan** on an opportunity does the same.
+- **Hide from the plan**: the row leaves the list and the counts. Undo it from the message that appears, or later from the hidden rows under the list, with **Show again**.
+
+Only admins and managers can mark priorities and hide rows.
+
+## What AI cites in your topics
+
+A line above the list shows what visitors from AI assistants did on your site in the period: visits and, with Google Analytics connected, key events, orders and revenue.
+
+Under **What AI cites in your topics**, click **Show**. For each page type, it compares the share of AI citations that go to that type with its share of your own cited pages. A type AI cites 10 points or more beyond your own mix is marked as a **Gap**. Click a row to filter the priority list to that page type; click it again, or **Clear**, to remove the filter. The page types follow your brand's **Site type** in the brand settings.
+
+## Shopping opportunities
+
+For an online shop whose product catalog, in **Settings › Integrations**, has categories, **Shopping opportunities** above the list shows two kinds of rows:
+
+- **Buying guide · New**: a category with at least three products in stock that no page of yours guides shoppers to yet. **Create** starts a buying guide draft and opens it in the editor.
+- **Existing post · Fix**: a published post that links a product that no longer sells, or misses some of its category pages. **Optimize** opens it in the editor.
diff --git a/help/content/content-quality-rules.md b/help/content/content-quality-rules.md
new file mode 100644
index 000000000..37c4881a0
--- /dev/null
+++ b/help/content/content-quality-rules.md
@@ -0,0 +1,42 @@
+---
+title: Write to the Content Editor rules
+route: /dashboard/content/editor
+keywords: [content rules, run agent, agent suggestions, accept suggestions, rules tab]
+synonyms: [writing rules, style guide rules, fix open rules, apply rules with agent]
+plan: all
+updated: 2026-10-29
+last_checked: 2026-10-29
+owner: Cihan
+app_version: 0.3.0
+approved: false
+sources: [docs/guides/content-editor.mdx]
+---
+Every page in the Content Editor is written to a set of rules for its page type. The **Rules** chip in the header shows how many the page meets now, and the Content Agent can suggest changes for the rules it doesn't meet yet.
+
+## The Rules tab
+
+In the side panel, **Instructions** opens on **Rules**. The rules come in five groups: page type rules, length and structure, fan-outs and keywords, links and sources, and style. Most rules can be turned on or off, and many have a value you can change. Each rule's status is read from the page as you write. Rules marked **AI check** are judged by the Content Agent and aren't counted yet.
+
+## Save or reset your rules
+
+Changes to the rules apply to this draft only. **Save as my rules** keeps them for your workspace's next drafts of this page type, and **Reset** goes back to the default set. Under **Your text rules**, write a rule in your own words, or use **Paste many** to add several from a style guide.
+
+## Let the agent fix open rules
+
+**Run Agent**, in the Rules tab, asks the Content Agent to suggest changes for the rules the page doesn't meet yet. Its suggestions appear in the page, each with **Accept** and **Reject**, under a bar with **Accept all** and **Reject all**.
+
+A decided suggestion can be taken back with **Undo** or **Restore**, and **Undo all** brings every one back to review. One run makes at most eight changes: run it again for the rules still open. Running the agent uses your workspace's monthly AI usage.
+
+## Before you run it again
+
+- Accept or reject the open suggestions before you run the agent again or edit the page as HTML.
+- The agent doesn't change heading levels: turn extra H1s into H2s in the toolbar yourself.
+- The agent only changes drafts, not published pages.
+
+## How your choices teach the agent
+
+What you accept and reject teaches the agent. For a rule whose suggestions this brand has mostly rejected over the last 90 days (at least five decided, on two pages or more, seven in ten rejected), it keeps its changes as small as the rule allows, and its card says which rules while it works. Only its small changes count: a rejected **Rewrite the whole page** doesn't.
+
+## On a page that is already live
+
+When the draft optimizes a live page, the agent is also told how that page does now, as the **Live page** tab shows it over 30 days: how often AI answers cite it and the prompts it lost. It keeps the answers and facts AI engines cite, answers a lost prompt's question where a rule's change can, and its card says what it kept in mind.
diff --git a/help/content/send-to-cms.md b/help/content/send-to-cms.md
new file mode 100644
index 000000000..0dbb62094
--- /dev/null
+++ b/help/content/send-to-cms.md
@@ -0,0 +1,61 @@
+---
+title: Send a page to your CMS
+route: /dashboard/content/library
+keywords: [send to cms, webflow, wordpress, framer, draft post, last send]
+synonyms: [push to cms, publish to blog, send to website, export to cms]
+plan: all
+updated: 2026-10-29
+last_checked: 2026-10-29
+owner: Cihan
+app_version: 0.3.0
+approved: false
+sources: [docs/guides/send-to-cms.mdx, docs/guides/content-editor.mdx]
+---
+**Send to CMS** sends a Content Editor page to your CMS: into a Webflow or Framer collection as a draft item, to your WordPress site or your Wix site's Blog as a draft post, or to your workflow tool, which creates the post in your CMS. Ansvisor never publishes on its own: you publish in your CMS. Only admins and managers can send pages.
+
+## Send one page
+
+In the editor, click the paper-plane icon in the header. **Before you publish** opens first when it has something to ask about, such as a missing author, open placeholders, missing alt text or broken links. Check it and click **Continue to send**. From the Library, open a row's **…** menu and pick **Send to CMS**.
+
+The dialog fills the **Slug**, **Meta title** and **Meta description** from the page; change them before sending if you like. What you send is kept on the draft, so the next send starts from it.
+
+## Choose where it goes
+
+The dialog offers each CMS that is connected and ready, Webflow, WordPress, Framer, then Wix, beside **Your workflow**. Your choice is remembered for the brand in your browser.
+
+A CMS that is connected but not ready shows what to fix and can't be picked, for example when its sign-in is refused or no collection is picked. If the dialog can't check a CMS you chose just then, the page goes to your workflow, and the dialog says so with **Try again**.
+
+## What holds a send
+
+The page goes as Export makes it, without its H1: your CMS's title field carries the title. Ansvisor holds the send while:
+
+- a placeholder in square brackets is still open. **Fill now** selects the first one;
+- the page has no title, or nothing under its title;
+- the slug can't be used.
+
+## Send several pages
+
+In the Library, tick the rows to send, up to 25 at once, and click **Send to CMS** in the bar above the table. The dialog lists each page as **Ready**, or says why it can't go; a page that can't go is left out, so fix it in the editor. The pages go one after another in the background. You can close the dialog at any time: the Library keeps following them.
+
+## See how a send went
+
+Beside each row's status, the Library shows the last send: a clock while it's on its way, a paper plane once it's sent, a red warning when it failed (hover it to see why), and a blue link once it's published (click it to open the live page).
+
+The **Last send** filter beside **Created** lists the pages that were **Never sent**, **Sending**, **Sent**, **Failed** or **Published**. A send to your workflow isn't retried on its own: send a failed page again from the Library.
+
+## When a page becomes Published
+
+Once you publish the item in Webflow or Framer, or the post in WordPress or Wix, the page becomes **Published** and its address shows in the Library. Ansvisor checks when you open the Library and once a day, for sends from the last 60 days. A page sent to your workflow becomes Published when the workflow reports where it went live. A page whose item or post you deleted reads as removed: send it again to make a new one.
+
+## Sending a page again
+
+Sending a page again updates the same Webflow or Framer item, the same WordPress draft post, or the same Wix post (a published one gets a new draft version in Wix). Changes to a published Webflow or Framer item go live with your next publish there. WordPress has no draft of a published post, so a page whose post is published, scheduled or private goes as a new draft post: move the text over in WordPress.
+
+## What each CMS leaves out
+
+- **Webflow**: tables go as lists, and images aren't sent. An empty meta description goes as the page's opening words.
+- **WordPress**: the meta title and the structured data aren't sent, so set them in your SEO plugin. Images keep their addresses, and no featured image or category is set.
+- **Framer**: tables go as tables, and images aren't sent. Another item in the collection with the same slug stops the send.
+- **Wix**: images aren't sent, and no cover image or category is set. Sending again replaces the post's text, so add images in Wix after your last send.
+
+Add images in your CMS. The slug may also change slightly there: lowercase, without accents, with dots turned into dashes.
diff --git a/help/content/start-new-content.md b/help/content/start-new-content.md
new file mode 100644
index 000000000..a3f0c556b
--- /dev/null
+++ b/help/content/start-new-content.md
@@ -0,0 +1,46 @@
+---
+title: Start a page in New content
+route: /dashboard/content/editor
+keywords: [new content, create page, optimize page, content agent, import page]
+synonyms: [write a new article, start a draft, new page, improve an existing page]
+plan: all
+updated: 2026-10-29
+last_checked: 2026-10-29
+owner: Cihan
+app_version: 0.3.0
+approved: false
+sources: [docs/guides/content-editor.mdx]
+---
+**New content** starts a page, either a new one or one you already have, written by you, the Content Agent or a teammate. It opens from the Plan and the Library, from a Plan row's button, from **Send to Editor** on an opportunity, and from **Create content** in a prompt's menu on Tracked Prompts. **Editor** in the sidebar opens it too when you have no draft to go back to.
+
+## Create a new page
+
+Pick **Create from scratch** to write a new page from:
+
+- a **Content Opportunity**;
+- a **Prompt**. Prompts where AI already names or cites you in the last 30 days aren't listed, as they need no new page;
+- a **Topic** that has tracked prompts.
+
+## Optimize a page you have
+
+Pick **Optimize content** to work on an existing page:
+
+- **Page URL**: a page on your brand's site. **Import** brings in its main text, headings, links and schema; menus and footers stay out.
+- **Paste text** from Word, Google Docs or a web page. Headings, lists, links and tables are kept.
+- **Upload file**: Word, Markdown, HTML or text, up to 10 MB.
+
+## Pick the page type
+
+Click **Continue**. Ansvisor picks a **Page type** from the title, the content, the page or the opportunity, and says which. Pick another if it's wrong; you can also change it later in the editor.
+
+For a new page, **Overview** lists your pages AI already cites for the same prompts, with how many of those prompts each one answers, its AI citations and its Google position. You can choose to optimize one of those pages instead, or **Create a new page**.
+
+## Choose who does it
+
+- **Me**, the default: **Open editor** opens the draft for you to write.
+- **Agent**: **Start the agent**, and the Content Agent writes a new page, or suggests changes to an existing one, in the background. The editor shows it at work. You can leave: the draft waits in the Library under **In review**. It uses this month's AI usage.
+- **Team**: **Send to Action Center** saves the draft in the Library and sends it to the Action Center as an action. Pick a teammate in **Assign to**, or leave it unassigned.
+
+## If the agent can't start
+
+If this month's AI usage is spent, or the agent is already writing another of your team's drafts, the agent doesn't start. The draft stays in the Library for you to write yourself.
diff --git a/help/getting-started/add-your-first-brand.md b/help/getting-started/add-your-first-brand.md
new file mode 100644
index 000000000..cb1cbf9d8
--- /dev/null
+++ b/help/getting-started/add-your-first-brand.md
@@ -0,0 +1,48 @@
+---
+title: Add your first brand
+route: /dashboard/brands/new
+keywords: [add brand, onboarding, setup wizard, first brand, brand setup]
+synonyms: [get started, create a brand, set up my account, onboard]
+plan: all
+updated: 2026-10-29
+last_checked: 2026-10-29
+owner: Cihan
+app_version: 0.3.0
+approved: false
+sources: [docs/guides/add-your-first-brand.mdx]
+---
+When you start with Ansvisor, a setup wizard walks you through your first brand. It takes 5 to 10 minutes, and you need the website of the brand you want to track. The wizard has five steps on a self-hosted install and seven on the cloud. Each step saves as you go, so you can leave and pick up where you stopped.
+
+## Brand basics
+
+- **Brand name**: exactly how the brand is named in the wild, not an abbreviation. Suggestions copy your spelling; mention detection ignores upper and lower case.
+- **Website**: your main domain, such as example.com. It's used to classify citations, attribute AI traffic and research competitors.
+- **Describe your brand** (optional, but recommended): one to three sentences on what you do and who your audience is. AI uses it to suggest better topics, prompts and competitors.
+
+There's no separate industry field: Ansvisor infers it from your description.
+
+## Market and language
+
+On **Select your target market**, pick the primary **Region** and **Language** for tracking. They decide which language AI engines answer in, the region weighting on engines that localize their answers, such as Google AI Overviews and AI Mode, and the search data used to estimate prompt volumes. You can change these later for each prompt if you track several regions.
+
+## Topics
+
+Ansvisor uses your description and an AI model to suggest topics: themes your customers care about. Check the topics that fit your brand and uncheck any that don't apply. You can also type a topic of your own and add it.
+
+Aim for four to seven topics. Too few gives a narrow view; too many is noisy and hard to read.
+
+## Prompts
+
+For each topic you chose, Ansvisor writes starter prompts: real questions a potential customer would ask an AI engine. Review the list: edit a prompt's text inline if the phrasing feels off, delete prompts that don't fit, and add your own at the bottom. The prompts run on every AI platform your plan includes.
+
+Keep prompts free of your brand name. A question about your brand by name almost always scores high, because the AI talks about you when asked, so it doesn't measure how AI engines find you on their own.
+
+## Competitors
+
+Ansvisor suggests competitors for your brand. Check the real rivals, uncheck any that aren't comparable, and add your own at the top, with the name AI answers use and, if you like, its domain. Every plan includes competitor tracking.
+
+## The last steps on the cloud
+
+**What do you work on?** asks what you work on or lead, and your left menu keeps only the sections you need. **Skip: show me everything** keeps every section, and you can change it later in **Settings › Account**. A self-hosted install asks the same the first time you open the dashboard.
+
+The final step picks your plan and starts checkout. Paid plans include a free trial: you enter card details, but you aren't charged until the trial ends. A self-hosted install skips this step.
diff --git a/help/getting-started/first-results-after-setup.md b/help/getting-started/first-results-after-setup.md
new file mode 100644
index 000000000..5d4ebc88e
--- /dev/null
+++ b/help/getting-started/first-results-after-setup.md
@@ -0,0 +1,35 @@
+---
+title: Your first results after brand setup
+route: /dashboard/insights
+keywords: [first results, after onboarding, first run, visibility zero, brand name prompts]
+synonyms: [no data yet, empty dashboard, why is my score zero, what happens next]
+plan: all
+updated: 2026-10-29
+last_checked: 2026-10-29
+owner: Cihan
+app_version: 0.3.0
+approved: false
+sources: [docs/guides/add-your-first-brand.mdx]
+---
+When you finish setting up a brand, the first round of questions to the AI engines starts on its own. How long it takes depends on how many prompts you track, and the dashboard updates as the AI platforms report back.
+
+## Where results appear first
+
+- **Brand Visibility**: the first metric cards appear soon after the run starts.
+- **Citations**: fills in as AI answers come back with the pages they cite.
+
+## My visibility score is 0
+
+Probably nothing is broken. New, niche or B2B brands often start at 0 on most prompts. AI engines need either plenty of training data that mentions your brand, or live citations from trusted sources. The score stays at 0 only while both are missing, and that is exactly what working on your AI visibility changes.
+
+## Prompts that include my brand name
+
+Remove them. A question that asks about your brand by name almost always scores high, because the AI obviously talks about you when asked. Such prompts don't measure how AI engines discover you on their own, and they distort your averages.
+
+## Competitors I've never heard of
+
+Unknown brands in AI answers are useful: they're the alternatives people see when you aren't cited. Add the most-mentioned ones to your competitors to track them properly.
+
+## Add more topics and prompts later
+
+You can add topics and prompts at any time, from the **Topics** tab of the **Prompts** page or from **Prompt Suggestions**. Tracking picks them up on the next scheduled run.
diff --git a/help/mentions/brand-perception.md b/help/mentions/brand-perception.md
new file mode 100644
index 000000000..1398b4a8a
--- /dev/null
+++ b/help/mentions/brand-perception.md
@@ -0,0 +1,30 @@
+---
+title: Brand perception in AI answers
+route: /dashboard/mentions
+keywords: [perception, brand attributes, brand shape, biggest gap, associated words]
+synonyms: [how AI describes my brand, brand themes, what AI says about us, words AI uses for us]
+plan: all
+updated: 2026-10-29
+last_checked: 2026-10-29
+owner: Cihan
+app_version: 0.3.0
+approved: false
+sources: [docs/guides/mentions.mdx]
+---
+The **Perception** tab of the Mentions & Sentiment page shows the words AI uses next to your brand, and how they compare with the words it uses for your competitors. The terms are taken from the sentences that name each brand by fixed rules, not by an AI model. A sentence that names two brands lends its words to both.
+
+## What AI associates with your brand
+
+The first card lists the words and phrases that appear next to your brand, ranked by the number of responses they appear in, each with its change from the previous period. Hover over a count to see its share of the responses naming you.
+
+## Brand shape and Brand × attribute
+
+**Brand shape: you vs top competitors** is a radar chart of your brand and your three most-mentioned competitors, on the attributes that tell them apart. It needs at least three attributes and at least one competitor named in the period. **Brand × attribute** shows the same attributes as a heat map: each attribute's share of the responses naming each brand, for you and up to four competitors.
+
+## Biggest gap
+
+Under the brand shape, **Biggest gap** names the attribute where a competitor leads you by the widest margin, when that margin is at least 20 points. For example, AI may use *affordable* in 45% of the responses naming a competitor but in only 10% of the responses naming you. Figures marked **Low sample** are left out.
+
+## Frequently mentioned with
+
+This card lists the competitors named in the same responses as you, with the change from the previous period. It shows whether the set of brands AI considers alongside you is changing.
diff --git a/help/mentions/correct-mention-sentiment.md b/help/mentions/correct-mention-sentiment.md
new file mode 100644
index 000000000..2a4520b66
--- /dev/null
+++ b/help/mentions/correct-mention-sentiment.md
@@ -0,0 +1,26 @@
+---
+title: Correct a wrong mention sentiment
+route: /dashboard/mentions
+keywords: [sentiment is wrong, correct sentiment, sentiment correction, wrong tone]
+synonyms: [fix sentiment, change sentiment, answer marked negative by mistake, misclassified mention]
+plan: all
+updated: 2026-10-29
+last_checked: 2026-10-29
+owner: Cihan
+app_version: 0.3.0
+approved: false
+sources: [docs/guides/mentions.mdx]
+---
+Every answer that names your brand is tagged Positive, Neutral or Negative toward your brand. If an answer's tone is tagged wrongly, you can correct it from the answer's side panel on the Mentions & Sentiment page, and every figure then counts it with the new sentiment.
+
+## Find and open the answer
+
+On the **Mentions** tab, the table's **Sentiment** filter or the **Most negative** sort helps you find the answers to check. Click a row to open the answer in a side panel. It shows the platform and model, the country, the date, the prompt with **View prompt**, the sentiment, your position and the topic. **Mention** shows the passage around your brand, **Also named in this answer** the competitors, **Full AI response** the answer as tracking stored it, and **Citations in this answer** each cited source.
+
+## Save a correction
+
+In the panel, click **Sentiment is wrong?**. Pick the new sentiment, add a note on why if you like, and click **Save correction**. From then on, every figure counts the answer with its new sentiment, and the original is kept in the history. The panel shows who made the correction, when, what the sentiment was before, and the note.
+
+## Who can correct a sentiment
+
+Only admins and managers can correct a sentiment. Other members can see past corrections in the panel, but not the **Sentiment is wrong?** button.
diff --git a/help/mentions/mention-metrics.md b/help/mentions/mention-metrics.md
new file mode 100644
index 000000000..91def84e8
--- /dev/null
+++ b/help/mentions/mention-metrics.md
@@ -0,0 +1,26 @@
+---
+title: How mention and sentiment figures are counted
+route: /dashboard/mentions
+keywords: [total mentions, mention rate, net sentiment, low sample, position, prompts with mentions]
+synonyms: [what does net sentiment mean, times named, mention count, sentiment score]
+plan: all
+updated: 2026-10-29
+last_checked: 2026-10-29
+owner: Cihan
+app_version: 0.3.0
+approved: false
+sources: [docs/guides/mentions.mdx]
+---
+Each tracked answer that names your brand is tagged **Positive**, **Neutral** or **Negative** toward your brand. An answer that doesn't name you has no sentiment. Competitors have no sentiment yet, so every tone on the Mentions & Sentiment page is the tone toward your brand.
+
+## Total mentions and mentions
+
+**Total mentions** counts every time an answer names your brand, by its name or one of its domains: an answer that names you three times counts three. The platform table's Mentions column and the exported file count the same way. A mention in the **Mentions** table is different: it is an answer that names you at least once, so an answer naming you twice is still one mention. **Competitor framing** counts responses the same way.
+
+## Rates, sentiment and position
+
+**Mention rate** is the responses naming you divided by the tracked responses in the period. **Prompts with mentions** counts the prompts with at least one answer naming you, out of the prompts tracked. **Net sentiment** is the positive share minus the negative share of the responses naming you, from −100 to +100. Neutral answers count in the total. It is shown as a signed number such as +21, not as a percentage. **Position** is your brand's place among the brands an answer names, 1 meaning named first.
+
+## Low sample
+
+A figure that rests on fewer than 20 responses is marked **Low sample**. For the mention rate that means fewer than 20 tracked responses; for net sentiment, fewer than 20 responses naming you. A low-sample figure is shown without a change from the previous period, isn't tinted in the platform table, and is left out of **Biggest gap**. With so few answers, a single negative response moves the number a lot, so widen the date range or clear a filter before you draw conclusions from it.
diff --git a/help/mentions/mentions-and-sentiment.md b/help/mentions/mentions-and-sentiment.md
new file mode 100644
index 000000000..7111f2953
--- /dev/null
+++ b/help/mentions/mentions-and-sentiment.md
@@ -0,0 +1,34 @@
+---
+title: Mentions & Sentiment page
+route: /dashboard/mentions
+keywords: [mentions, sentiment, brand mentions, mention rate, net sentiment, co-mentioned]
+synonyms: [where AI names my brand, tone of AI answers, brand mentions in AI answers, how AI talks about us]
+plan: all
+updated: 2026-10-29
+last_checked: 2026-10-29
+owner: Cihan
+app_version: 0.3.0
+approved: false
+sources: [docs/guides/mentions.mdx]
+---
+The **Mentions & Sentiment** page, **Mentions** in the sidebar, shows where AI answers name your brand, in what tone, and next to which competitors. It reads the answers your tracking already stores, so opening it or changing a filter never asks an AI model again. It has three tabs: **Overview**, **Mentions** and **Perception**.
+
+## Filters
+
+All three tabs share one filter bar, for the brand active in the sidebar. **Date range** offers 24h, 7d, 30d (the default), 90d or Custom; a custom range is read as whole UTC days. **Platform** picks one AI platform or all of them. **Country** appears only when the brand tracks more than one country, and **Topic** only when it has topics. **Search** matches the prompts' text, also on the Perception tab, where the box reads *Search keywords…*. The tab and the filters are kept in the page's address, so a refresh or a shared link opens the same view.
+
+## Overview tab
+
+The **Overview** opens with a summary line: net sentiment and its change, the platform behind most of that change, and the competitors AI usually names next to you. Four cards follow: **Total mentions**, **Mention rate**, **Prompts with mentions**, and **Net sentiment** with its positive, neutral and negative split. **Mentions & sentiment over time** draws one point per day; the 24h view has no trend. **By AI platform** compares the platforms and notes when two are 40 or more points apart. **Competitor framing** shows how often AI names each competitor and, under **Also names you**, how many of those answers name you too.
+
+## Changes from the previous period
+
+Every change compares the selected period with the period of the same length just before it. It is taken only on the prompts that were answered in both periods, so adding or pausing prompts doesn't show up as a rise or a fall.
+
+## Mentions tab
+
+The **Mentions** tab lists every answer that names your brand, 100 per page. Each row shows the prompt, the platform, the **Mention context** around your brand, the **Sentiment**, the **Position** (#2 of 5 means named second of five brands), the **Co-mentioned** competitors, the topic and the date. The table's own controls filter by **Sentiment** and **Co-mentioned**, **Group by** prompt or platform, and **Sort** by Newest, Oldest, Most co-mentions or Most negative. Click a row to open the full answer in a side panel.
+
+## Copy summary and Export CSV
+
+The page header has three buttons. **Copy summary** copies the Overview as plain text: the filters, the summary line, the four figures and the period, ready to paste into a doc or a ticket. **Export CSV** downloads the answers that name your brand under the filters, one row per answer, up to 5,000 rows per file. From the Mentions tab, the file also follows the table's sentiment and co-mention filters and its sort. Lines above the rows record the filters, the period and the counts. **Reports** opens the Reports page.
diff --git a/help/mentions/send-mentions-to-action-center.md b/help/mentions/send-mentions-to-action-center.md
new file mode 100644
index 000000000..33cbc3d10
--- /dev/null
+++ b/help/mentions/send-mentions-to-action-center.md
@@ -0,0 +1,26 @@
+---
+title: Act on mentions with the Action Center and content
+route: /dashboard/mentions
+keywords: [send to action center, suggest content, act on mentions, negative mentions]
+synonyms: [improve how AI describes us, fix negative answers, turn mentions into tasks, create action from mentions]
+plan: all
+updated: 2026-10-29
+last_checked: 2026-10-29
+owner: Cihan
+app_version: 0.3.0
+approved: false
+sources: [docs/guides/mentions.mdx]
+---
+From the **Mentions** tab of the Mentions & Sentiment page, you can turn answers that name your brand into work: one Action Center action for a set of answers, or a content suggestion for an answer's prompt.
+
+## Send answers to the Action Center
+
+Tick the rows you want; the box in the table header selects every answer on the page. A bar appears with **Send to Action Center**, **Export** for a file of just the selected rows, and **Clear**. You can also send a single answer from its side panel. One send takes up to 100 answers, a full page. Sending the same answers again doesn't create a second action: the notification says **Already in the Action Center** and opens the existing one.
+
+## What the action contains
+
+The selected answers become one action, titled *Improve how AI describes you for "…"*, or *across N prompts* when they cover several. It quotes up to 10 of the answers with their platform, sentiment and excerpt. It is a **Protect** action when any of the answers is negative, and a **Growth** action otherwise; the larger the share of negative answers, the higher its impact. Its tasks walk you from inspecting the contexts you're mentioned in and analyzing the cited sources to strengthening how your brand is described, publishing, and measuring the outcome on the answers' prompts.
+
+## Suggest content
+
+**Suggest content** in an answer's side panel writes a content suggestion for the answer's prompt. When it is ready, the notification opens it in Content. If the prompt already has a suggestion, **Open existing** takes you there instead. Suggestions count toward your monthly content limit. A prompt can't get a suggestion until it has tracked answers and a prompt group; both arrive with its next tracking run.
diff --git a/help/prompts/add-prompts.md b/help/prompts/add-prompts.md
new file mode 100644
index 000000000..26c0db9a9
--- /dev/null
+++ b/help/prompts/add-prompts.md
@@ -0,0 +1,34 @@
+---
+title: Add prompts to track
+route: /dashboard/prompts/tracked
+keywords: [add prompt, bulk import, track prompts, new prompt]
+synonyms: [add a question, add queries, track a search]
+plan: all
+updated: 2026-10-29
+last_checked: 2026-10-29
+owner: Cihan
+app_version: 0.3.0
+approved: false
+sources: [docs/guides/prompts.mdx]
+---
+A prompt is a question Ansvisor asks the AI engines for your brand every day, such as "best project management tool for agencies". Its answers show where your brand is named and cited. You can add prompts one at a time, many at once, or from suggestions.
+
+## Add one prompt
+
+On **Prompts › Tracked Prompts**, click **Add prompt**. On the **Single** tab, type the question and pick its topic, platforms and AI models, then save. It is tracked in the brand's location; add more later with **Track in more locations**.
+
+## Add many at once
+
+In the same dialog, open the **Bulk import** tab and upload a one-column Excel or CSV file of prompts. A preview marks each row as ready, a duplicate, an error or over your plan limit, and the rows that fit are imported. Set their topic and location afterwards from each prompt's **Edit**.
+
+## Add from suggestions
+
+Switch the list to **Prompt Suggestions**. Add one suggestion with its **+** button, or every listed one with **Add All**.
+
+## When the first results come
+
+A prompt has no results until it is first asked. **Add prompt** and **Bulk import** start right away when they can. A prompt added from suggestions, or one whose analysis couldn't start, gets its first results with the next daily run. Until then, for its first two days, its row says **Within 24h**.
+
+## What counts on your plan
+
+Every prompt counts on your plan's prompt limit, once in each location it is tracked in. Paused prompts don't count. If adding would pass the limit, Ansvisor adds the prompts that fit and says how many weren't added. Once the limit is reached, it adds nothing until you pause or remove prompts.
diff --git a/help/prompts/ai-search-volume.md b/help/prompts/ai-search-volume.md
new file mode 100644
index 000000000..c098492b6
--- /dev/null
+++ b/help/prompts/ai-search-volume.md
@@ -0,0 +1,54 @@
+---
+title: How estimated AI search volume works
+route: /dashboard/prompts/insights
+keywords: [ai search volume, estimate, confidence, volume range, estimate badge]
+synonyms: [how many people ask, monthly AI searches, AI volume, est. ai volume]
+plan: all
+updated: 2026-10-29
+last_checked: 2026-10-29
+owner: Cihan
+app_version: 0.3.0
+approved: false
+sources: [docs/guides/prompts.mdx]
+---
+AI assistants don't publish how often people ask them something, so every AI search volume in Ansvisor is an estimate. It is shown as a range, such as 540–2k, never as a single number, and carries a dashed **Estimate** badge.
+
+## How it's estimated
+
+The estimate is the median monthly Google searches for the prompt's related keywords, times the share of Google searches that now get an AI answer (45% by default; a self-hosted server can change it).
+
+When a prompt's volume is analyzed, Ansvisor picks five short Google keywords for its category and reads their monthly Google searches. Keywords Google has no data for are left out, and the median of the rest is the prompt's **Google Vol.** The Insights tab shows the AI share as **AI Adoption Rate**.
+
+## How the range is set
+
+The range is the AI share of the middle half of the keywords, kept within half to twice the estimate and at least 25% either side of it. With only one keyword to go by, the range is half to twice the estimate.
+
+## Confidence levels
+
+- **High**: at least 3 of the prompt's keywords have Google search data, and the middle half of them are within 4× of each other.
+- **Medium**: at least 2 have data, within 15× of each other.
+- **Low**: anything else, including an estimate of 0.
+
+Volumes fetched over six months ago count one level less; **Refresh Volumes** on the Insights tab gets newer ones.
+
+## Low-confidence estimates
+
+A low-confidence estimate is grey, its badge reads **Estimate · low confidence**, and it isn't used to rank on its own. **High Opportunity Prompts** leaves it out, and sorting **Tracked Prompts** by **Volume** puts it after the confident estimates, before prompts with no volume.
+
+## Read a range
+
+Hover or focus a range or its badge to open the method note. It says how the estimate was built, how many of the related keywords have Google search data, the sources and their month, and the confidence with its date.
+
+Measured numbers never carry the badge: your visibility, mentions and citations, Search Console clicks and impressions, or a suggestion's Analytics sessions. If a number has no **Estimate** badge, it was counted, not estimated.
+
+## Totals, single keywords and AI guesses
+
+Some estimates are built differently, and their note says so:
+
+- **Totals and averages** (**Total Est. AI Volume**, **Avg. AI Volume**, and Similar Topics' **Est. AI Volume**): the sum of the lows to the sum of the highs. A total is high-confidence when at least 80% of its volume is, low when more than half is low-confidence, and medium otherwise. The note gives the mix.
+- **One keyword** (each row of Similar Topics): the Google volume is measured and only the AI share is assumed, so it is Medium with a ±25% range.
+- **An AI's guess** (suggestions from Ansvisor Agents, and the Brand Visibility page's Opportunities): always Low, from half to twice the guess. Track the prompt to get an estimate from its keywords' Google searches.
+
+## Compare prompts by their ranges
+
+Read volumes as ranges: compare prompts by their confidence as well as their size, and prefer high-confidence estimates when you decide what to work on first.
diff --git a/help/prompts/create-content-from-prompt.md b/help/prompts/create-content-from-prompt.md
new file mode 100644
index 000000000..0a0b13e52
--- /dev/null
+++ b/help/prompts/create-content-from-prompt.md
@@ -0,0 +1,44 @@
+---
+title: Create content from a tracked prompt
+route: /dashboard/prompts/tracked
+keywords: [create content, optimize content, improve position, open draft, content for a prompt]
+synonyms: [write a page for a prompt, improve ranking, start a draft from a prompt]
+plan: all
+updated: 2026-10-29
+last_checked: 2026-10-29
+owner: Cihan
+app_version: 0.3.0
+approved: false
+sources: [docs/guides/prompts.mdx]
+---
+Each row in **Tracked Prompts** has a **…** menu with **Create content** and **Improve position…**. Use them to start a page written for the prompt, or to improve a page of yours that AI engines already cite.
+
+## Start a page with Create content
+
+1. Click **Create content**. It opens **New content** at its Start step, titled with the prompt.
+2. Pick a page type. Ansvisor picks one from the title; change it if it's wrong, now or later in the editor.
+3. Check the overlap. Start reads your pages AI already cites for the prompt. If one already answers it, you can optimize that page instead of choosing **Create a new page**.
+4. Choose who does it: **Me** opens the Content Editor, **Agent** writes a draft for you to review, and **Team** sends it to the Action Center.
+
+## When a draft already exists
+
+If a draft that isn't published yet already covers the prompt, one started from the prompt, its topic or an opportunity, the menu shows **Open draft** instead. That way you don't start a second page for the same prompt.
+
+## The content step in Improve
+
+**Improve position…** opens the prompt's positions panel. Its **Improve** tab starts with a content step, decided from the answers in the panel's period:
+
+- No answer mentions or cites your brand: **Create content**, to start a new page.
+- Your brand is named in under 20% of the answers, with 5 or more answers: **Optimize content**.
+- Your average position is worse than #3, over 3 or more answers that name you: **Optimize content**.
+- Anything else: no content step.
+
+**Optimize content** opens Start on your page AI engines cite most for the prompt. When none is cited, Start lists your pages that answer the prompt so you can pick one. When a draft already covers the prompt, the step offers **Open draft**.
+
+## How Improve's suggestions are written
+
+The **Improve** tab's suggestions per platform are written from the answers' evidence and from this brand's last two Improve actions on the prompt for the same region: which of their suggestions were done or skipped, and how each platform moved since. They build on what moved, and don't suggest again what was done on a platform that didn't move, or what your team skipped. An action sent less than a week ago is too early to tell either way.
+
+## Suggestions that use competitor pages
+
+When the Action Center's **Analyze competitor content** has read the pages cited for the prompt in the last month, the **Improve** tab's suggestions also see what Site Audit's checks found on them: which checks each competitor page passes, and which they pass that your cited page doesn't.
diff --git a/help/prompts/prompt-detail.md b/help/prompts/prompt-detail.md
new file mode 100644
index 000000000..72c5a6583
--- /dev/null
+++ b/help/prompts/prompt-detail.md
@@ -0,0 +1,56 @@
+---
+title: A prompt's own page and its real demand
+route: /dashboard/prompts/tracked
+keywords: [prompt page, prompt detail, demand, google trends, people also ask, search console]
+synonyms: [single prompt view, is this asked for real, real demand, prompt results]
+plan: all
+updated: 2026-10-29
+last_checked: 2026-10-29
+owner: Cihan
+app_version: 0.3.0
+approved: false
+sources: [docs/guides/prompts.mdx]
+---
+Click a prompt's text in **Tracked Prompts** to open its own page. It shows the prompt's visibility, mentions, citations and per-platform results. When the prompt has a volume, an **AI searches a month** card shows its estimate. Beside it, **Demand · is this asked for real?** shows measured signals of whether people really ask it.
+
+## The four demand signals
+
+Each signal shows its source and the day it was read:
+
+- **Search Console**: impressions and clicks over the last 28 days for your Google queries that match the prompt, how many matched and the most seen one. Read every night.
+- **Google Trends**: a 12-month line for the prompt's keyword, **Rising**, **Steady** or **Falling**, and the change on a year ago. Read about once a week.
+- **People Also Ask**: the related questions Google shows for the keyword. Read about once a week.
+- **YouTube**: the views across YouTube's top videos for the keyword, and the most viewed one. Read about once a week.
+
+A query matches a prompt when most of its words are in the prompt.
+
+## When a signal is missing
+
+Google Trends, People Also Ask and YouTube are read for the keyword with the most Google searches in the prompt's volume analysis, so the prompt needs its volume analyzed first. These weekly reads are off by default, and a self-hosted server can turn them on; until then their cards say **Weekly reads aren't turned on yet**. Search Console is read every night either way.
+
+A card without its signal says why, for example **Connect Search Console to see real clicks** (with a **Connect Search Console** button for admins and managers), the first sync still running, or **Analyze volume first**.
+
+## Track a People Also Ask question
+
+Each People Also Ask question has a **+ Track** button. One click adds it as a tracked prompt in the prompt's topic, within your plan's prompt limit. A question your brand already tracks, however it's written, shows **Tracked** instead and isn't added twice.
+
+## Demand level
+
+The signals read come to a demand level, shown in the prompt's KPI row, for example **Demand · High** with **3 of 4 signals**. It carries a **Measured** badge, never **Estimate**.
+
+## How the demand level is scored
+
+Each signal read gives 2 points when it's strong, 1 when there is some, 0 otherwise:
+
+- **Search Console**: 1,000 impressions or more is strong, 100 or more is some.
+- **Google Trends**: rising is strong, steady is some.
+- **People Also Ask**: 4 questions or more is strong, 1 or more is some.
+- **YouTube**: 100,000 views or more is strong, 5,000 or more is some.
+
+The level is **High** at 60% of the possible points or more with at least two signals read, **Medium** at 30% or more, and **Low** below. A signal that was read and found nothing counts; one never read doesn't.
+
+## Demand and the Priority Score
+
+The demand level feeds the Priority Score of the Action Center's actions about the prompt. **High** raises its demand to at least what about 2,500 monthly AI searches would give, and **Medium** to about 160. **Low** leaves it as it is, because signals that found little don't prove nobody asks.
+
+When a prompt's level changes, the brand's open actions are scored again that night, and **Why this priority?** names a High or Medium level.
diff --git a/help/prompts/prompt-insights.md b/help/prompts/prompt-insights.md
new file mode 100644
index 000000000..50901c6ee
--- /dev/null
+++ b/help/prompts/prompt-insights.md
@@ -0,0 +1,45 @@
+---
+title: Prompt Insights tab and volume analysis
+route: /dashboard/prompts/insights
+keywords: [prompt insights, analyze volumes, refresh volumes, intent, similar topics, high opportunity prompts]
+synonyms: [prompt volume, search demand, keyword volume, volume analysis]
+plan: all
+updated: 2026-10-29
+last_checked: 2026-10-29
+owner: Cihan
+app_version: 0.3.0
+approved: false
+sources: [docs/guides/prompts.mdx]
+---
+The **Insights** tab of the **Prompts** page shows the estimated AI search volume of your tracked prompts, their intent and the highest-demand prompts. It has no date range, because volume estimates aren't tied to a period.
+
+## Run a volume analysis
+
+Volumes come from an analysis you run:
+
+- **Analyze Volumes** the first time;
+- **Analyze New Prompts** for prompts added since;
+- **Refresh Volumes** for newer Google data;
+- **Re-analyze Keywords** to analyze every prompt again, keywords included.
+
+On cloud plans, each analysis counts against a monthly quota, and the toolbar shows how many are left. **Tracked Prompts** also warns when some prompts haven't been analyzed yet.
+
+## What the tab shows
+
+For analyzed prompts you see:
+
+- **Total Est. AI Volume** and **Avg. AI Volume** across your prompts.
+- The **Prompt Volumes** table, with each prompt's **Google Vol.** and **Est. AI Vol.**
+- Each prompt's **Intent**: Comparison, How-to, What is, Best / Top, vs. / Review, Recommendation, Problem Solving or Other.
+- **High Opportunity Prompts**: the prompts with the most estimated AI searches. A low-confidence estimate isn't listed there on its own.
+- **Similar Topics**: the most-searched keyword clusters across your prompts. **See all** opens its own page.
+- **AI Adoption Rate**: the share of Google searches that get an AI answer, used in every estimate.
+
+## Where else estimated volumes show
+
+- **Tracked Prompts**: the **Volume** column.
+- A prompt's own page: an **AI searches a month** card, when the prompt has a volume.
+- **Prompt Suggestions**: each AI suggestion's monthly range.
+- Outside Prompts: the Action Center's affected prompts and topic opportunities, the Brand Visibility page's Opportunities, and a content opportunity's **Est. AI Volume**.
+
+Exports mark estimated volumes, with their range and confidence, so a spreadsheet doesn't read them as counts.
diff --git a/help/prompts/prompt-locations.md b/help/prompts/prompt-locations.md
new file mode 100644
index 000000000..193dcab96
--- /dev/null
+++ b/help/prompts/prompt-locations.md
@@ -0,0 +1,67 @@
+---
+title: Track prompts in more locations
+route: /dashboard/prompts/tracked
+keywords: [locations, add location, remove location, locations view, countries, us states, cities]
+synonyms: [track in another country, regional tracking, geo tracking, stop tracking a location]
+plan: all
+updated: 2026-10-29
+last_checked: 2026-10-29
+owner: Cihan
+app_version: 0.3.0
+approved: false
+sources: [docs/guides/prompts.mdx]
+---
+A prompt can be tracked in several locations (countries, US states or cities), with its own figures in each one. In **Tracked Prompts** you compare them in the **Locations** view, and add or remove places with the **Track in more locations** dialog.
+
+## The Locations view
+
+**View: Table | Locations**, above the table, switches to **Locations**: one row per prompt and one column per location your prompts track, the most-tracked first (12 at most). Pick **Visibility**, **Position**, **Mentions** or **Citations**; cells are shaded light to dark, and for Position, named earlier is darker.
+
+An **All prompts · average** row and a **Gap** column (the best location minus the worst) sit beside three cards: **Weakest location**, **Strongest location** and **Biggest gap between locations**. Click a cell to open the prompt's Positions panel for that location. The date range and filters work as in the table.
+
+## A location as a table column
+
+To see one location in the normal table, open **Columns** and use **Add a location column**. It adds a metric in one location, such as Visibility in one country, as its own column.
+
+## Add locations
+
+Open **Track in more locations** from **+ Add location** in the Locations view, **+ Track** in a cell, or **Locations › Add locations** in the bar that shows when you select prompts. Then:
+
+1. Pick one or more countries, US states or cities. A country one engine doesn't answer from says so, and that engine is skipped there while the others run.
+2. Choose the prompts: the one you clicked, the selected ones, a topic, or every prompt shown.
+3. Check **Plan usage**. Every prompt in every location counts once on your plan's prompt locations. Paused prompts don't count.
+
+## When locations would pass your plan's limit
+
+If the choice would pass your plan's limit, the dialog says how many it needs and how many are left, and saves nothing. Pick fewer prompts or places, or follow your plan's way to track more.
+
+## Which engines answer from a state or city
+
+**US states** are asked by every engine from the state itself. Google's engines asked a state country-wide until October 2026, so their answers from before then stay in the country's column.
+
+**Cities** are asked only by the engines that can answer from a city: Google AI Mode, Google AI Overview and the API models (OpenAI, Claude). ChatGPT, Perplexity, Copilot, Gemini and Grok answer per country, so they skip a city and their answers stay in the country's column; track the country too to see them. A city column says **AI Mode · AIO · APIs**. A city counts as one location on your plan, like a country.
+
+## What a new location shows
+
+A location you just added has no answers until the next daily run asks it. Until then, its cells say:
+
+- **Within 24h**: it's tracked there, and the first results come with the next daily run.
+- **Paused**: the prompt is paused, so it isn't asked anywhere until you resume it.
+- **No engine here**: none of the prompt's engines answers from that place. Add an engine that does, or remove the location.
+
+When your plan's prompt locations are used up, **+ Track** reads **Limit reached**. It still opens the dialog, which shows your plan's way to track more.
+
+## Stop tracking in a location
+
+Open the same dialog in **Remove** mode from:
+
+- the **…** menu on a location's column in the Locations view, then **Stop tracking in** that location;
+- the **×** on a tracked cell (**Stop tracking here**), when the prompt is tracked somewhere else too;
+- **Locations › Remove locations** in the bar for selected prompts;
+- **Remove** at the top of **Track in more locations**.
+
+Pick the locations and the prompts. A prompt needs at least one location, so one tracked only in the places you stop keeps them.
+
+## What happens to past results
+
+Past results stay. The prompt is no longer asked there from the next daily run, and the answers already collected there aren't deleted. The dialog shows how many tracked locations it removes and how many it frees on your plan; a paused prompt's locations go too, but free nothing. When no prompt shown is tracked in a location any more, its column leaves the Locations view.
diff --git a/help/prompts/prompt-suggestions.md b/help/prompts/prompt-suggestions.md
new file mode 100644
index 000000000..595d3b69b
--- /dev/null
+++ b/help/prompts/prompt-suggestions.md
@@ -0,0 +1,53 @@
+---
+title: Prompt Suggestions and Add All
+route: /dashboard/prompts/suggestions
+keywords: [prompt suggestions, add all, clear all, suggested prompts, source filter]
+synonyms: [prompt ideas, recommended prompts, new prompt ideas, suggest queries]
+plan: all
+updated: 2026-10-29
+last_checked: 2026-10-29
+owner: Cihan
+app_version: 0.3.0
+approved: false
+sources: [docs/guides/prompts.mdx]
+---
+**Prompt Suggestions** gives you new prompts to track. Open it with the switch above the filters on the **Prompts** tab of the **Prompts** page. Each refresh writes up to 25 prompt ideas, each with a topic label and a one-line reason.
+
+## Where suggestions come from
+
+Suggestions come from four sources, and the **Source** filter shows one at a time:
+
+- **Ansvisor Agents**: written by AI from your brand and tracking data. Each shows an estimated monthly AI volume as a grey range.
+- **Google Search Console**: queries people already search, with their impressions a month.
+- **Google Analytics (GA4)**: pages that earn revenue or conversions, or already get AI visits, with that evidence on the row.
+- **Your catalog**: for an online shop with a product catalog, the questions a shopper asks before buying in its categories, using their price bands. Hover the badge for the category; the **Shopping** filter lists them.
+
+When Search Console or Analytics is connected, a share of each refresh is drawn from their data.
+
+## Add, dismiss or refresh suggestions
+
+Use **Search prompts…** to narrow the list, **+** to add a suggestion, **×** to dismiss it and **New** to generate a fresh list. **Clear All** asks first, then dismisses every suggestion listed. Those from Search Console or Analytics stay hidden for 30 days once dismissed.
+
+Suggestions expire after 48 hours. Click **New** for fresh ideas, for example after you change your content.
+
+A prompt added from suggestions gets its first results with the next daily run.
+
+## Add every suggestion at once
+
+**Add All** adds every suggestion listed to **Tracked Prompts**, and the added ones leave the list. Like **Clear All**, it acts on what's listed, so a source filter or a search narrows it. Each prompt passes the same plan check as a single **+**.
+
+## When your prompt limit stops Add All
+
+When your plan's prompt limit stops **Add All** partway, an amber note above the list says how many suggestions couldn't be added and how many were added before the limit.
+
+To add the rest, upgrade your plan (or contact us to raise your limit on Growth and Enterprise), or pause as many prompts as the note names and try again. **Go to Tracked Prompts** opens the list where you can pause prompts. Self-hosted installs have no prompt limit, so everything listed is added.
+
+## How AI suggestions are written
+
+The suggestion engine considers five signals:
+
+1. Your brand's name, description and language.
+2. Your tracked prompts, so it doesn't repeat them.
+3. Your topics, so it reuses their labels when relevant.
+4. The competitors cited most in your recent tracked AI answers.
+5. The estimated volumes of your tracked prompts, so its own estimates stay in proportion.
diff --git a/help/prompts/query-fan-outs.md b/help/prompts/query-fan-outs.md
new file mode 100644
index 000000000..013ad813a
--- /dev/null
+++ b/help/prompts/query-fan-outs.md
@@ -0,0 +1,22 @@
+---
+title: Query Fan-outs behind AI answers
+route: /dashboard/prompts/fan-outs
+keywords: [query fan-out, fan-outs, searches behind answers, copilot, perplexity]
+synonyms: [sub-queries, background searches, AI search queries, what the AI searched]
+plan: all
+updated: 2026-10-29
+last_checked: 2026-10-29
+owner: Cihan
+app_version: 0.3.0
+approved: false
+sources: [docs/guides/prompts.mdx]
+---
+Behind an answer, an AI engine can run searches of its own. The **Query Fan-out** tab of the **Prompts** page lists the searches the AI engines ran behind their answers to your tracked prompts. In the sidebar, the **Prompts** item opens it as **Query Fan-outs**.
+
+## Which engines share their searches
+
+Copilot and Perplexity share the searches behind their answers. ChatGPT's are sampled on some days, and the ChatGPT API shares them when it searches. Google AI Overviews, AI Mode and Gemini don't share them, so those engines can't appear on this tab.
+
+## Pick the date range
+
+The tab uses the **Prompts** page's date range: 24h, 7d, 30d, 90d or custom. The range applies to Prompts, Topics and Query Fan-out alike, and carries over when you switch between these tabs.
diff --git a/help/prompts/topics.md b/help/prompts/topics.md
new file mode 100644
index 000000000..a643eecab
--- /dev/null
+++ b/help/prompts/topics.md
@@ -0,0 +1,40 @@
+---
+title: Topics tab for prompts grouped by theme
+route: /dashboard/prompts/topics
+keywords: [topics, topic, add topic, delete topic, topic visibility, top competitor]
+synonyms: [themes, prompt groups, clusters, categories]
+plan: all
+updated: 2026-10-29
+last_checked: 2026-10-29
+owner: Cihan
+app_version: 0.3.0
+approved: false
+sources: [docs/guides/topics.mdx, docs/guides/prompts.mdx, docs/guides/add-your-first-brand.mdx]
+---
+**Topics** is a tab of the **Prompts** page. It groups your tracked prompts by theme, so you can see how your brand does in AI answers at a higher level than single prompts.
+
+## What each topic shows
+
+Each row shows the topic's name and how many prompts it has, the average visibility across its prompts, its **Trend** against the window of the same length just before, and the **Top competitor** within the topic.
+
+The tab uses the page's date range (24h, 7d, 30d, 90d or custom) and opens on the last 24 hours. It has no Location selector, because topics aren't split by location.
+
+## Open a topic
+
+Click a topic to open its details: how each of its prompts performs, and how often competitors are mentioned within the topic.
+
+## Add a topic
+
+Use **Add Topic** on the tab to add a new topic at any time. You can also add prompts at any time, for example from **Prompt Suggestions**. Tracking picks them up on the next scheduled run.
+
+Keep topic names short (two to four words) and intent-driven: "Pricing" works better than a long descriptive phrase.
+
+## Delete a topic
+
+Admins and managers open the **…** menu at the end of a topic's row and choose **Delete topic**. The topic's prompts aren't deleted: they keep their tracking and results, as prompts without a topic. You can undo the delete for 10 seconds.
+
+## Use topics for planning
+
+- **Editorial planning**: pick the topic with the lowest score and the weakest competitor lead; that's your next content sprint.
+- **Reporting**: topic-level stories, such as which themes you own and which you're losing, are easier to share than prompt-level metrics.
+- **Spotting drift**: a topic that drops while others stay flat usually signals one competitor's surge or content that has aged.
diff --git a/help/prompts/tracked-prompts.md b/help/prompts/tracked-prompts.md
new file mode 100644
index 000000000..d3daab7ed
--- /dev/null
+++ b/help/prompts/tracked-prompts.md
@@ -0,0 +1,55 @@
+---
+title: Tracked Prompts list, statuses and columns
+route: /dashboard/prompts/tracked
+keywords: [tracked prompts, prompt status, columns, performance filter, paused prompts]
+synonyms: [prompt list, my prompts, prompt table, tracked queries]
+plan: all
+updated: 2026-10-29
+last_checked: 2026-10-29
+owner: Cihan
+app_version: 0.3.0
+approved: false
+sources: [docs/guides/prompts.mdx]
+---
+**Tracked Prompts** lists every prompt Ansvisor asks the AI engines for your brand, with its results over the date range you pick (24h, 7d, 30d, 90d or custom). It sits on the **Prompts** tab of the **Prompts** page: a switch above the filters picks **Tracked Prompts** or **Prompt Suggestions**. The date range carries over when you switch to Topics or Query Fan-out.
+
+## Prompt statuses
+
+Each row has a status:
+
+- **New** (amber): added in the last 24 hours. New prompts lead the table until you sort by a column; after a day they read as Active.
+- **Active**: tracked on its schedule.
+- **Paused**: kept, but not tracked.
+
+The table opens on **Active + New**. The **Status** filter, or the cards above the table, show one status or all of them.
+
+## Rows without figures
+
+A new prompt has no results until it's first asked. Until then, for its first two days, the row says **Within 24h** in its first figure column, such as Visibility. A paused prompt says **Paused** there when the Status column is hidden. An older prompt with no results in the range, such as days without a run, shows a dash, as does a range that ended before today.
+
+## Choose your columns
+
+**Columns**, beside the search box, picks the table's columns: tick one to show or hide it, and drag it, or use its arrows, to move it. **Prompt** always shows. The **Added** column shows when each prompt was added; click it to sort by it. The **Volume** column shows each prompt's estimated AI search volume.
+
+Your columns are saved for you on this brand; nobody else's table changes.
+
+## Share a column view with the brand
+
+Admins and managers can tick **Share with everyone on** the brand: their columns become the brand's view, which everyone without columns of their own sees. Anyone can still change their own. **Reset to the brand's view** goes back to the brand's columns, and **Reset to default** to the default columns when the brand has none.
+
+## Filter by performance
+
+The **Performance** filter groups prompts by where your brand is named in their AI answers over the selected range, across every platform the prompt tracks:
+
+- **Best**: named in the first third of the brands an answer lists.
+- **Good**: named in the middle third.
+- **Late**: named in the last third.
+- **Missing**: named in fewer than half the answers.
+
+Late and Missing list the furthest behind first. A prompt with no answers in the range, such as a new one, has no performance yet. The dots in the **Positions** column show each platform on its own.
+
+## Open a prompt or act on it
+
+Click a prompt's text to open its own page, with its visibility, mentions, citations and per-platform results. Each row's **…** menu offers **Edit**, **Pause tracking**, **Improve position…**, **Suggest content**, **Create content** and **Delete**.
+
+**Export CSV** downloads the list. Estimated volumes in it are marked as estimates, with their confidence and range, so a spreadsheet doesn't read them as counts.
diff --git a/help/reports/daily-pulse.md b/help/reports/daily-pulse.md
new file mode 100644
index 000000000..5f5ed6cd8
--- /dev/null
+++ b/help/reports/daily-pulse.md
@@ -0,0 +1,59 @@
+---
+title: Daily Pulse email digest
+route: /dashboard/settings/notifications
+keywords: [daily pulse, digest, notifications, email summary, highlights, warnings]
+synonyms: [daily email, weekly email, brand digest, visibility alerts, stop the emails]
+plan: all
+updated: 2026-10-29
+last_checked: 2026-10-29
+owner: Cihan
+app_version: 0.3.0
+approved: false
+sources: [docs/guides/daily-pulse.mdx]
+---
+The Daily Pulse is a short digest for each brand that goes out after its scheduled tracking run. It shows the headline numbers for the period, then lists what went well (highlights) and what needs attention (warnings). It comes by email on Ansvisor Cloud, and can also go to your webhook, on cloud and self-hosted alike.
+
+## Choose how often it comes
+
+Open **Settings › Notifications**. Every active brand has its own row with a frequency and a recipient list. Admins and managers can change them; other members see them read-only. The **Manage notification settings** link at the bottom of every pulse email opens this page.
+
+- **Daily** (the default): after every scheduled daily tracking run.
+- **Weekly**: once a week on Monday (UTC), covering the last 7 days compared with the 7 days before.
+- **Only when notable**: like Daily, but skipped when there are no highlights and no warnings.
+- **Off**: no email and no webhook.
+
+## Choose who gets it
+
+Enter email addresses separated by commas. Leave the field empty to send to every member of the organization. Each address is checked when you save: if one isn't valid, the save is rejected and the invalid entries are listed under the field, so fix or remove them and save again. Up to 50 addresses are stored. Recipients apply to the email only, not to the webhook.
+
+## When it is sent
+
+A pulse follows a brand's scheduled daily tracking run; manual runs and single-prompt refreshes never trigger one. It waits up to 90 minutes for late answers, so its numbers match the dashboard. No pulse is sent when the frequency is **Off**, or **Weekly** and it isn't Monday; when the brand already had one that day; when there are no tracking results in the window; when the frequency is **Only when notable** and nothing is notable; or, on the cloud, when the subscription isn't active or trialing.
+
+## What's in a pulse
+
+The KPIs are **Visibility**, the **7-day trend**, **Mentions** and **Citations**, **Positive sentiment**, and **Prompts you appear in**. For Daily and Only when notable, the period is the latest completed tracking run, at most 24 hours, compared with the run before it. For Weekly, it is the last 7 days compared with the 7 days before. The email subject carries the visibility score and the weekly trend; on Weekly, it says Weekly Pulse.
+
+## Highlights
+
+- **First citation**: a URL from a prompt's Target URLs list was cited by an AI engine for the first time. Up to 5 are listed.
+- **Prompt gain**: a prompt's visibility score rose by 5 points or more this week versus last week. The top 3 are listed.
+- **Competitor overtaken**: last week a competitor's score was at or above yours; this week yours is higher.
+- **New engine**: your brand was mentioned or cited on an AI platform for the first time ever.
+
+## Warnings
+
+- **Sharp drop**: your visibility score fell week over week by at least 3 points and at least 30% of last week's score, for brands with 10 or more tracked prompts.
+- **Competitor surge**: a competitor's score rose by 15 points or more week over week.
+- **Competitor crossed**: a competitor at or below your score last week is above it this week.
+- **Lost citations**: a prompt with an estimated AI volume above zero was cited on at least 60% of its result days in the last 14 days, then not at all on its 3 most recent.
+
+A smaller fall shows as **Visibility slipping**, in the email and the webhook alike.
+
+## Repeats and collection problems
+
+The same warning about the same subject, such as a surge by the same competitor, isn't repeated within 7 days. Highlights have no such pause. If an AI platform's result volume collapses on Ansvisor's side, the pulse treats it as a collection problem rather than a real change: **Sharp drop**, **Visibility slipping** and **Lost citations** are held back, and the email notes which platforms were affected.
+
+## Send it to a webhook
+
+Every pulse also goes to the brand's active webhook, if it is subscribed to the Daily Pulse event: under **Content**, open **Webhook Settings** and tick **Daily Pulse created**. On self-hosted installs, this is the only way the pulse is delivered. The webhook carries the same KPIs, highlights and warnings as the email. The frequency applies to it too; the recipient list doesn't.
diff --git a/help/reports/looker-studio.md b/help/reports/looker-studio.md
new file mode 100644
index 000000000..a8a5aa356
--- /dev/null
+++ b/help/reports/looker-studio.md
@@ -0,0 +1,41 @@
+---
+title: Connect Looker Studio
+route: /dashboard/settings/api-keys
+keywords: [Looker Studio, connector, GA4, blend data, metric sets]
+synonyms: [Google Data Studio, visibility in my own dashboard, connect a BI tool, reporting dashboard]
+plan: all
+updated: 2026-10-29
+last_checked: 2026-10-29
+owner: Cihan
+app_version: 0.3.0
+approved: false
+sources: [docs/guides/looker-studio.mdx]
+---
+The Ansvisor connector for Looker Studio pulls your AI visibility metrics into Looker Studio. One dashboard can then show AI visibility next to organic traffic, such as GA4 sessions and Search Console clicks, on one date axis. While the connector awaits Looker Studio's partner gallery, it is shared by link: ask Ansvisor for the current one.
+
+## Connect
+
+1. In **Settings › API Keys**, click **New key**, give it a name and click **Create**. Copy the key now: it is shown only once.
+2. Open the connector's link. Looker Studio asks you to authorize the connector, then asks for a key: paste the one you copied.
+3. Pick the brand and a metric set, then click **Connect** and **Add to report**.
+
+## Metric sets
+
+Each data source gives one metric set; add several data sources to use more than one in a report.
+
+- **Visibility Trend (daily)**: by date, the average visibility score, mentions, citations, tracked results and the competitor average.
+- **Share of Voice by Platform**: by platform, brand mentions, competitor mentions and share of voice.
+- **Citations — Top Domains**: by domain and category, citations, results citing and usage.
+- **AI Traffic by Platform**: visits by platform.
+
+The report's date range control sets the dates for every chart; no data source needs its own dates.
+
+## Blend with GA4
+
+Visibility Trend is built for blending: join it to a GA4 source on the **Date** dimension, and plot the average visibility score against sessions on a dual-axis time series. A rising visibility line that comes before a traffic lift is the clearest way to show AI search's impact to a client.
+
+## Good to know
+
+- Results are kept for five minutes per user, so a dashboard with many charts doesn't ask Ansvisor again for each one.
+- A key belongs to its creator's organization: only that organization's brands appear in the brand picker.
+- Revoking the key in **Settings › API Keys** breaks the data source at once, and Looker Studio asks for a new key.
diff --git a/help/reports/visibility-reports.md b/help/reports/visibility-reports.md
new file mode 100644
index 000000000..261abce05
--- /dev/null
+++ b/help/reports/visibility-reports.md
@@ -0,0 +1,37 @@
+---
+title: Generate visibility reports
+route: /dashboard/reports
+keywords: [reports, generate report, report templates, PDF, executive summary, snapshot]
+synonyms: [monthly report, client report, export a report, download PDF, report for my boss]
+plan: all
+updated: 2026-10-29
+last_checked: 2026-10-29
+owner: Cihan
+app_version: 0.3.0
+approved: false
+sources: [docs/guides/reports.mdx]
+---
+The **Reports** page makes point-in-time snapshots of your brand's AI visibility. Unlike the live dashboards, a report never changes: its numbers are captured when it is generated and stay the same as new tracking results arrive. That makes it the right tool for a monthly review, a client deliverable, or a before-and-after comparison around a campaign.
+
+## Generate a report
+
+Click **Generate Report** and pick a template. A template is a default set of sections, and you can switch any section on or off before generating. Then give the report a title if you like, pick the **Brand** and the **Date Range** (last 7, 30 or 90 days, or Custom), and check the **Sections**. Generated reports appear under **Generated Reports** on the same page, newest first.
+
+## Templates
+
+- **Weekly Visibility Summary**: KPIs, Trend and Share of Voice, over the last 7 days.
+- **Monthly Executive Report**: every section, over the last 30 days.
+- **Competitor Benchmark**: KPIs, Share of Voice, Competitors and Topics, over the last 30 days.
+- **Citation & Sources Report**: Citations and Citation Evidence, over the last 30 days.
+
+## Sections
+
+The full picker offers KPIs, Trend, Share of Voice, Competitors, Topics, Prompts, Mention Evidence, Fan-out, AI Traffic, Product Visibility, Audit Score, Citations and Citation Evidence. **Product Visibility** appears only for brands with Shopping mode on. The AI executive summary is always included: it isn't a section you pick.
+
+## Read a report
+
+A report opens with its **Executive Summary**: a short account of what changed versus the previous period, such as the headline visibility movement, mention and citation growth, and sentiment and competitor shifts. One card per selected section follows, in a fixed order. Each metric is compared with the previous period of equal length, so a 30-day report is compared with the 30 days before it. Percentage changes read +X%, and point changes for visibility and sentiment +X pts. Because the snapshot is frozen, a report's numbers can differ from the live dashboards.
+
+## Download a PDF
+
+**Download PDF** saves the report as a PDF with the same sections as the web view, the executive summary included. Its text stays selectable and sharp at any zoom, and the file is small enough to attach to an email.
diff --git a/help/settings/mcp-server.md b/help/settings/mcp-server.md
new file mode 100644
index 000000000..a3ac0bfff
--- /dev/null
+++ b/help/settings/mcp-server.md
@@ -0,0 +1,61 @@
+---
+title: Use Ansvisor in AI assistants with MCP
+route: /dashboard/settings/api-keys
+keywords: [mcp, mcp server, api key, claude, cursor, codex, connect assistant]
+synonyms: [model context protocol, use ansvisor in claude, ai assistant integration, connector]
+plan: all
+updated: 2026-10-29
+last_checked: 2026-10-29
+owner: Cihan
+app_version: 0.3.0
+approved: false
+sources: [docs/guides/mcp-server.mdx]
+---
+Ansvisor has an MCP (Model Context Protocol) server, so AI assistants such as Claude, Codex, Cursor, VS Code, Windsurf, Gemini CLI and Zed can query your Ansvisor data in a normal conversation. There is nothing to install: the assistant connects with an API key from **Settings › API Keys**.
+
+## Create an API key
+
+1. Open **Settings › API Keys**.
+2. Click **New key** and give it a name you'll recognize, such as "MCP — my laptop".
+3. Click **Create** and copy the key. It's shown only once, so store it somewhere safe, such as your password manager.
+
+## Connect your assistant
+
+The dialog that shows the key also has a ready-to-paste setup for each assistant under **Connect a client**, with the key filled in. Pick yours, click **Copy setup**, and follow the line under it, which says where the setup goes. Because it includes the key, copy it before you close the dialog.
+
+In Claude's desktop app, for example, you add Ansvisor under **Settings › Connectors › Add custom connector**. On Claude's Team and Enterprise plans, an owner adds the connector in the organization settings first.
+
+On a self-hosted install, the assistant connects to your own Ansvisor instead.
+
+## Check the connection
+
+Ask your assistant "List my Ansvisor brands". It should show your brands. If it reports a 401 error, the key is missing from the setup or was revoked.
+
+## What you can ask
+
+Ask in plain words, for example:
+
+- "What's my visibility score on ChatGPT for the last 7 days?"
+- "Which topic has the most prompts?"
+- "Who are my top 5 competitors by mention count this month?"
+- "Why did our sentiment drop on Gemini?"
+
+The assistant can read visibility, mentions, citations, prompts, AI traffic, saved reports, content, Site Audits, product visibility and the Action Center. Some requests change data, such as accepting a suggested prompt, writing a content draft or recording a task's progress. Assistants are told to make a change only when you ask for it.
+
+## What your key may do
+
+A key acts as the person who created it, with their role. Reading, drafts and suggestions in the Content Editor, Site Audits, task progress and an opportunity's progress are open to every role.
+
+Adding prompts or topics, reviewing tasks, dismissing, and work that costs AI, such as content briefs and Improve actions, are for admins and managers. Dismissing a prompt suggestion is open to analysts too. Anything a role can't do is refused before it changes anything.
+
+## Publish a draft from your assistant
+
+Ansvisor never sends a page to your CMS through MCP. Instead, the assistant can get a draft as a page ready for your CMS, put it there with its own CMS connector, and then mark the draft published at the address it went live at, so Ansvisor measures it.
+
+## When the plan has ended
+
+On Ansvisor Cloud, a workspace without an active plan is read-only. Its API keys keep working for reading, but every change is refused until an admin chooses a plan in **Settings › Billing**. Retrying doesn't help. New API keys can still be created; on a read-only workspace they only read. A self-hosted instance is never read-only.
+
+## Revoke a key
+
+If a key leaks or you stop using an assistant, open **Settings › API Keys** and click the trash icon next to the key. Assistants using it lose access immediately.
diff --git a/help/settings/menu-access.md b/help/settings/menu-access.md
new file mode 100644
index 000000000..bb8da919a
--- /dev/null
+++ b/help/settings/menu-access.md
@@ -0,0 +1,42 @@
+---
+title: Choose who sees which menu sections
+route: /dashboard/settings/menu-access
+keywords: [menu access, hide menu sections, role menu, customize menu, sidebar sections]
+synonyms: [hide pages from users, left menu per role, restrict menu, simplify sidebar]
+plan: all
+updated: 2026-10-29
+last_checked: 2026-10-29
+owner: Cihan
+app_version: 0.3.0
+approved: false
+sources: [docs/guides/menu-access.mdx]
+---
+**Menu access** in Settings decides which sections are on each person's left menu, so everyone sees the part of Ansvisor they use. It only hides: a section off someone's menu still opens from a link, and what their role can do there doesn't change. To change what someone can do, change their role in **Settings › Team**.
+
+## Who can change menus
+
+Admins change menus, and can let managers do it too: under **Who can change menus**, switch a manager on. That manager then sees **Menu access** in Settings, and nothing else of the admins' settings. They change what each role sees and every non-admin's menu, never an admin's.
+
+Only admins choose who changes menus. A manager who becomes another role stops changing menus.
+
+## Choose what each role sees
+
+**Sections each role sees** lists the left menu's sections, by group, with a column for each role. Tick what a role sees; each box is saved as you click it.
+
+- **Admin** always sees **Everything**.
+- **Action Center** is on everyone's menu and can't be turned off.
+- **Use a set…** fills a role's column from a work area's ready-made set, or **Everything**. Adjust it from there.
+
+Until an admin hides something, every role sees every section.
+
+## Change one person's menu
+
+**People** lists everyone but the admins, with their role, the work areas they picked, their menu and how many sections it keeps. A menu reads like **Analyst's menu**, or **Custom: +Prompts, −Content** when it differs from their role's.
+
+Click **Customize** to tick or untick a section for that person alone, even one their role hides. Each change is marked **shown for them** or **hidden for them**, and saved as you click. **Use Analyst's menu**, named after their role, takes their changes back.
+
+## How a menu is worked out
+
+A person's menu is their role's sections, changed by their own exceptions from **People**, then narrowed to the sections of their work areas, when they picked any.
+
+**Action Center** always stays, and nobody takes a section off an admin's menu; only an admin's own work areas narrow it. A section added to Ansvisor later is on everyone's menu until someone hides it.
diff --git a/help/settings/work-areas.md b/help/settings/work-areas.md
new file mode 100644
index 000000000..d39de04ac
--- /dev/null
+++ b/help/settings/work-areas.md
@@ -0,0 +1,42 @@
+---
+title: Pick your work areas
+route: /dashboard/settings/account
+keywords: [work areas, what do you work on, my menu, hidden sections, sections off your menu]
+synonyms: [missing menu item, where is a page, simplify my menu, my role focus]
+plan: all
+updated: 2026-10-29
+last_checked: 2026-10-29
+owner: Cihan
+app_version: 0.3.0
+approved: false
+sources: [docs/guides/menu-access.mdx]
+---
+Each person can say what they work on or lead. Their left menu then keeps only those areas' sections. Pick one or more; picking none keeps every section. **Settings** is always on the menu, with your API keys and MCP.
+
+## The work areas
+
+- **Founders & leaders**: Action Center, Chat, Brand Visibility, Mentions, AI Referral Traffic, Reports.
+- **AEO/GEO & Audit**: Action Center, Content, Audit, Citations, AI Referral Traffic, Reports.
+- **Content**: Action Center, Chat, Content, Prompts, Citations, Reports.
+- **Brand & PR**: Action Center, Chat, Brand Visibility, Mentions, Citations, Reports.
+- **Growth**: Action Center, Chat, Brand Visibility, Prompts, Citations, AI Referral Traffic, Reports.
+- **Product Development**: Action Center, Chat, Brand Visibility, Prompts, Mentions, Product Visibility, Reports.
+- **Business Development**: Action Center, Chat, Brand Visibility, Prompts, Mentions, Citations, Reports.
+- **Developers**: Action Center, Audit, Prompts, Citations, AI Referral Traffic, Reports.
+- **Analytics**: Action Center, Brand Visibility, Prompts, Mentions, Citations, Product Visibility, AI Referral Traffic, Reports.
+
+## When you're asked
+
+**What do you work on?** is asked once:
+
+- when you sign up, after your competitors and before you choose a plan;
+- when you join from an invitation, on your first sign-in;
+- on a self-hosted install, the first time the admin opens the dashboard.
+
+**Skip: show me everything** keeps every section.
+
+## Change your work areas
+
+Change your answer any time in **Settings › Account › Your work areas**. It also shows your menu, and why a section is off it: **Hidden by your admin** or **Not in your work areas**.
+
+When something is off your menu, the bottom of the sidebar says how much, such as **3 sections off your menu**, with **Change** leading to your work areas.
diff --git a/help/site-audit/fix-audit-issues.md b/help/site-audit/fix-audit-issues.md
new file mode 100644
index 000000000..a3d9add1b
--- /dev/null
+++ b/help/site-audit/fix-audit-issues.md
@@ -0,0 +1,50 @@
+---
+title: Fix site audit issues
+route: /dashboard/audit
+keywords: [ai recommendations, fix audit issues, placeholders, send to action center, re-audit]
+synonyms: [audit fixes, improve audit score, fix failed signals, raise page score]
+plan: all
+updated: 2026-10-29
+last_checked: 2026-10-29
+owner: Cihan
+app_version: 0.3.0
+approved: false
+sources: [docs/guides/site-audit.mdx]
+---
+An audit report doesn't only score a page: it writes fixes for what failed. Start with the **AI Recommendations**, then open the failing signals. When you've made your changes, click the re-run icon at the top of the report to audit the page again.
+
+## AI Recommendations
+
+**AI Recommendations**, at the top of the report, collects the fixes written for this page, ordered by priority: **high**, **medium** or **low**. Each names the signal and says what to change in one sentence. Where it helps, it gives a ready-to-paste draft, such as a meta description, an H1, an opening paragraph, FAQ questions and answers or sources to cite, and code, such as an FAQ schema block.
+
+Mechanical signals, such as HTTPS, the viewport tag, alt text or a video embed, keep their general fix.
+
+## Placeholders in brackets
+
+Recommendations only use facts found on the page, or your brand's name, description and domains when the page is on one of your domains. A number, percentage, price, date or company name found in neither becomes a placeholder in square brackets, highlighted in the draft, such as [customer name], [real metric], [price], [date] or [source URL].
+
+Replace each one with your real data before you publish. Search the pasted text for an opening bracket first: a placeholder left in schema code tells AI engines nothing.
+
+## Copy drafts and code
+
+Each draft and code block has a copy icon in its corner, **Copy draft** or **Copy code**. It copies the block exactly as shown, placeholders included, and a **Copied** message confirms it.
+
+## Send a page to the Action Center
+
+On a page that fails checks, **Send to Action Center** sends the page and the checks it fails as one Fix action, without waiting for the nightly pass. Sent again while that action is open, it doesn't add another: a notice names the open action, with **Open**. Only a page's latest report can be sent; an older one offers **Open the latest report**.
+
+Once you complete the action, it is judged by the page's next audit: the checks it failed now passing.
+
+## How audits feed the Action Center
+
+Each night the Action Center reads each page's latest audit from the last 90 days and turns failed signals into actions:
+
+- **Fix low-scoring pages**, for audited pages scoring below 50;
+- **Fix AI crawler accessibility**, **Fix structured data issues** and **Fix outdated content**;
+- **Fix weak entity signals**, **Fix internal content connectivity** and **Fix content gaps**.
+
+An action lists the pages it covers and the failed checks on each, with **Open Site Audit**.
+
+## Re-audit after you fix
+
+Only each page's latest audit counts. Once a new audit no longer fails a check, the Action Center stops reporting that page; an audit older than 90 days no longer counts either. Auto-audit also re-audits, in its next run, the pages whose Site Audit fix you completed in the Action Center.
diff --git a/help/site-audit/read-an-audit-report.md b/help/site-audit/read-an-audit-report.md
new file mode 100644
index 000000000..a1bb2f367
--- /dev/null
+++ b/help/site-audit/read-an-audit-report.md
@@ -0,0 +1,46 @@
+---
+title: Read a site audit report
+route: /dashboard/audit
+keywords: [audit report, audit score, signals, audit categories, grade]
+synonyms: [audit results, page score, what failed, E-E-A-T score]
+plan: all
+updated: 2026-10-29
+last_checked: 2026-10-29
+owner: Cihan
+app_version: 0.3.0
+approved: false
+sources: [docs/guides/site-audit.mdx]
+---
+Each audit has its own report. It opens right after you run an audit, or from a row on the **Pages** tab. It scores the page out of 100, breaks the score down by category and lists every signal checked. The icons at the top re-run the audit on the same page or delete it; **Delete audit** can't be undone.
+
+## Total score
+
+The gauge shows the page's score out of 100 and a grade: **Good** (70 and above), **Needs work** (40 to 69) or **Poor** (below 40). Beside it are the number of signals that **passed**, raised **warnings** or **failed**, and how many of the 48 signals were evaluated.
+
+## Categories
+
+**Category breakdown** scores each category out of 100, with how many of its signals were evaluated:
+
+- **Structure** (25%): whether an AI engine can parse the page, such as heading hierarchy, FAQ schema, lists and tables, and alt text.
+- **Content** (25%): whether the page answers its question directly and completely.
+- **Authority** (20%): whether it reads as expert and sourced, such as author byline, citations, statistics and freshness.
+- **E-E-A-T** (20%): experience, expertise and trust, such as case studies and author credentials.
+- **Trust** (10%): crawl-time basics such as AI bot access in robots.txt, llms.txt, HTTPS, canonical URL and meta description.
+
+## How the score is worked out
+
+A category's score is the average of its signals, and the total is the weighted average of the categories. A signal that couldn't be evaluated, shown as n/a, is left out, so it never lowers the score.
+
+Most signals are read straight from the page, its robots.txt and its llms.txt. Thirteen, such as **Direct answer to query** and **Case study evidence**, are judged by an AI model that first works out what the page itself is about, so a page is never judged against a topic it isn't trying to cover. **Brand entity in Wikidata** looks your brand up in Wikidata.
+
+## Signals
+
+The **Signals** section lists every check by category, each with a pass, warning, fail or n/a icon and its impact: **high**, **medium** or **standard**. **Issues only** hides the passing ones. Open a signal to see:
+
+- the **Finding**: what this audit saw on the page, when the check reports it;
+- what the signal measures, and **Why** it matters to AI engines;
+- the **Recommended fix** written for this page, or the general **Fix** for signals without one.
+
+## Product pages
+
+On a product page, one with Product structured data, **Product description opening** checks that the description's first 40 words, or 250 characters, name the product type and at least two key attributes, such as a measure, a material, a fit or a use. ChatGPT appears to read only about that much of a description when it matches products to a shopping question. The finding shows where the opening ends. Other pages show this signal as n/a.
diff --git a/help/site-audit/run-a-site-audit.md b/help/site-audit/run-a-site-audit.md
new file mode 100644
index 000000000..2152fa5da
--- /dev/null
+++ b/help/site-audit/run-a-site-audit.md
@@ -0,0 +1,55 @@
+---
+title: Run a site audit
+route: /dashboard/audit
+keywords: [site audit, run audit, audit a page, auto-audit, audit settings]
+synonyms: [page audit, check my page, scan a page, AI readiness check]
+plan: all
+updated: 2026-10-29
+last_checked: 2026-10-29
+owner: Cihan
+app_version: 0.3.0
+approved: false
+sources: [docs/guides/site-audit.mdx]
+---
+**Site Audit** checks how ready a page is to be cited by AI answer engines. It fetches the page, runs 48 signals in five categories, scores the page out of 100 and writes fixes for what failed. You find it in the sidebar under **Execution**, after **Content**. It is on every plan.
+
+## Start an audit
+
+The field at the top of the Site Audit page is filled in with your brand's primary domain. Replace it with the page you want to check and click **Run Audit**.
+
+You can audit any page: your own, or a competitor's page to see what it does better. Audits of other sites are listed with your pages but kept out of your site's score.
+
+## While it runs
+
+The report opens right away and follows the audit through its stages: **Fetching the page…**, **Analyzing 48 signals…** and **Generating AI recommendations…**. An audit usually takes about 30 seconds. If it is still running after two minutes, the page says **This audit is taking longer than expected.** Click **Check again** to look again.
+
+## Other ways to start an audit
+
+- **Audit selected** on the **Opportunities** tab;
+- **Re-audit** on the pages you select on the **Pages** tab;
+- **Audit page** on a landing page of the **AI Referral Traffic** page;
+- the re-run icon at the top of a report, to audit the same page again;
+- auto-audit, which audits your top-priority pages on its own.
+
+Audits started in bulk run a few at a time.
+
+## Your monthly audits
+
+Each completed audit uses one of your plan's monthly audits. An audit still running already counts; a failed one doesn't. The **Audits this month** card shows what's left. Self-hosted instances have no monthly limit.
+
+## How auto-audit works
+
+Auto-audit spends a quarter of your monthly audits on the top pages of the **Opportunities** tab, in five runs on days 1, 7, 14, 21 and 28 of your billing month.
+
+Pages you fixed go first: a page whose Site Audit fix you completed in the Action Center in the last 30 days, and that hasn't been audited since, so the action's result says whether the fix took. Pinned pages come next. Pages audited in the last 30 days wait for a later run. After a run, a notice sums it up with **View pages**, and its audits show **Auto** in the Pages list.
+
+## Site Audit settings
+
+**Settings**, at the top right of the Site Audit page, holds:
+
+- **Auto-audit**: on or off for every brand in your organization. Turned off, all your monthly audits are yours to spend.
+- **E-mail the team after each run**: a summary e-mail after each auto-audit run.
+- **Excluded pages**: one path per line, where an asterisk stands for anything. These pages stay out of Opportunities and auto-audits.
+- **Pinned pages**: the pages you pinned from Opportunities, with **Unpin**.
+
+Only admins and managers can change these settings; other members see them read-only.
diff --git a/help/site-audit/site-audit-overview.md b/help/site-audit/site-audit-overview.md
new file mode 100644
index 000000000..c0768878e
--- /dev/null
+++ b/help/site-audit/site-audit-overview.md
@@ -0,0 +1,53 @@
+---
+title: Site Audit scores, opportunities and pages
+route: /dashboard/audit
+keywords: [site score, pages audited, audit opportunities, audited pages, score over time]
+synonyms: [audit overview, audit history, which pages to audit, audit priority]
+plan: all
+updated: 2026-10-29
+last_checked: 2026-10-29
+owner: Cihan
+app_version: 0.3.0
+approved: false
+sources: [docs/guides/site-audit.mdx]
+---
+The **Site Audit** page sums up your audits: a summary line, four cards and a chart, then two tabs. **Opportunities** lists the pages worth auditing next, and **Pages** lists every page you audited.
+
+## Summary and cards
+
+A summary line gives your site score, pages audited, lowest category and next auto-audit. The first three cards count only the pages of your primary domain:
+
+- **Site score**: the average score of each page's latest audit, and its change from each page's previous audit.
+- **Pages audited**: pages of your domain with a completed audit, and how many pages of other sites you audited.
+- **Authority · E-E-A-T**: those two category scores averaged over your pages, and the signals that fail or warn on the most pages.
+- **Audits this month**: audits left this billing month, with the share auto-audit keeps for its next runs.
+
+## Score over time
+
+**Score over time** charts your primary domain's audit scores, one point a day from the latest audit that day, over 7d, 30d, 90d or All. The category breakdown of the latest audit sits beside it. The chart needs audits on at least two days.
+
+## Why a page is in Opportunities
+
+The **Opportunities** tab is an inbox of pages worth auditing next. Each row shows a **Priority** from 0 to 100, the page, **Why** it's listed, its **AI visits** over the last 30 days and its **Last audit**. A page is listed when:
+
+- **No AI traffic**: it ranks high on your site in Google Analytics but gets no visitors from AI engines. **Targeted, not cited** adds that a tracked prompt points at it and no AI answer cites it;
+- a competitor is cited on prompts where you never are, and this is your closest page;
+- it was **Never audited**, or not in the last 90 days.
+
+AI visits need Google Analytics or the tracking snippet.
+
+## Priority and templates
+
+The priority multiplies the page's value to the business, the size of the opportunity and how easy it is to fix, meaning how close its last score is to passing. Hover the number to see the three parts.
+
+Pages built from one template, such as every post of a blog, share most of their gaps. When at least two-thirds of a template's audited pages fail the same signals, the template gets one row: **Fix once, applies to** its pages. Auditing it audits the template's pages that have no audit yet.
+
+## Select, pin or exclude
+
+Select rows to **Audit selected**, to **Pin** them, so they always come first, or to **Exclude** them. Excluded pages are listed under the table, where **Restore** puts them back.
+
+## The Pages tab
+
+**Pages** lists every page you audited, one row per page, most recent first, with its **Score**, its **Issues** (signals that failed or warned), its **Status** and its **Last audit**. Status is **Good** (70 or more), **Improving** (below 70 but up from the previous audit), **Needs work**, **Running** or **Failed**. **Auto** marks an auto-audit's audit.
+
+Click a row to open its latest report. The arrow beside a page lists its earlier audits, each with **View report** and a delete icon. **Group by template** groups the rows under their template. **Export** downloads the list as CSV or Excel; select pages to **Re-audit** them or export only those.
diff --git a/help/visibility/brand-visibility.md b/help/visibility/brand-visibility.md
new file mode 100644
index 000000000..34556e63c
--- /dev/null
+++ b/help/visibility/brand-visibility.md
@@ -0,0 +1,52 @@
+---
+title: Brand Visibility page overview
+route: /dashboard/insights
+keywords: [brand visibility, answer engine insights, visibility score, trend chart, filters, opportunities]
+synonyms: [AI visibility report, analytics hub, insights page, dashboard overview]
+plan: all
+updated: 2026-10-29
+last_checked: 2026-10-29
+owner: Cihan
+app_version: 0.3.0
+approved: false
+sources: [docs/guides/answer-engine-insights.mdx, docs/guides/prompts.mdx]
+---
+**Brand Visibility** in the sidebar opens **Answer Engine Insights**, the main analytics page for a brand. It shows how AI engines represent your brand across your tracked prompts, and where to act next.
+
+## What the page shows
+
+- Key metric cards at the top: your visibility score, mentions, citations and tracked prompts, each with its change.
+- A trend chart of your visibility, to spot dips and match them to content you published or to competitor activity.
+- **By AI platform**: which AI engines do best and worst for your brand. Use it to route content work toward your weakest platform.
+- **Topics vs competitors**: your topics against your competitors, shaded light to dark.
+
+## Filters
+
+At the top of the page you can narrow everything to:
+
+- **AI Model**: one AI engine;
+- **Region**: one region, when your prompts run in several;
+- **Topic**: one topic;
+- a date range.
+
+The filters apply to the whole page at once.
+
+## Opportunities
+
+**Opportunities** suggests prompts to track next. **Track** on a suggestion opens **Tracked Prompts** in a new tab, with the new prompts on top.
+
+The volumes shown with these suggestions are an AI's guess, since no keyword behind them is measured yet, so they are always low confidence: grey, from half to twice the guess. Track a prompt to get an estimate from its keywords' Google searches.
+
+## A weekly review
+
+1. Check the brand-level change. If visibility is down by 10% or more, use the trend chart to find when the drop happened.
+2. Find the AI platform contributing least: that's your next investment.
+3. Look at the suggested prompts and pick one to three to add.
+4. Start one piece of content for the prompt that did worst this week.
+
+## When a metric drops sharply
+
+1. Read the AI answers for what dropped. Is your brand still mentioned but ranked lower, or completely absent?
+2. Look at the competitors those answers name. A new competitor surge is often the cause.
+3. Compare the cited pages with last week's. A retracted article or a new competitor post is often the trigger.
+4. Decide what to do: publish content aimed at the prompt, earn a placement on a site AI engines cite, or accept it as a one-time dip.
diff --git a/help/visibility/visibility-metrics.md b/help/visibility/visibility-metrics.md
new file mode 100644
index 000000000..c6bc2852a
--- /dev/null
+++ b/help/visibility/visibility-metrics.md
@@ -0,0 +1,52 @@
+---
+title: What visibility, position, mentions and citations mean
+route: /dashboard/insights
+keywords: [visibility score, position, mentions, citations, performance, measured metrics]
+synonyms: [metrics explained, what is visibility, ranking in AI answers, why is my score zero]
+plan: all
+updated: 2026-10-29
+last_checked: 2026-10-29
+owner: Cihan
+app_version: 0.3.0
+approved: false
+sources: [docs/guides/answer-engine-insights.mdx, docs/guides/prompts.mdx, docs/guides/add-your-first-brand.mdx]
+---
+Ansvisor describes how AI answers treat your brand with a few metrics: visibility, position, mentions and citations. They are counted from real AI answers, not estimated, so they never carry an **Estimate** badge.
+
+## Read your visibility score
+
+Visibility is your brand's score on the **Brand Visibility** page, and a column in **Tracked Prompts**. A rough guide:
+
+- **Under 15**: cold start. AI engines barely know the brand on these prompts.
+- **15 to 35**: emerging presence. AI sometimes cites you, but competitors dominate.
+- **35 to 60**: established. You're reliably cited; focus on the platforms or topics still trailing.
+- **Over 60**: category leader. Your job is defending your place against new entrants.
+
+The right threshold depends on how competitive the category is: in a crowded category, 20 can already be leadership.
+
+## Watch the change, not the number
+
+Don't fixate on the absolute number; watch how it moves. A brand growing from 18 to 24 in a quarter is healthier than one stuck at 50.
+
+A score of 0 doesn't mean something is broken. New, niche or B2B brands often start at 0 on most prompts. AI engines need either plenty of training data that mentions your brand or live citations from trusted sources, and the score stays at 0 only while both are missing.
+
+## Position and Performance
+
+Position is where your brand is named among the brands an AI answer lists; named earlier is better. The **Performance** filter in **Tracked Prompts** groups prompts by it:
+
+- **Best**: named in the first third of the brands listed.
+- **Good**: in the middle third.
+- **Late**: in the last third.
+- **Missing**: named in fewer than half the answers.
+
+The dots in the **Positions** column show each platform on its own.
+
+## Mentions and citations
+
+A mention is an AI answer naming your brand. Mention detection ignores upper and lower case, but suggestions copy your spelling, so name the brand as it's named in the wild.
+
+A citation is a source page that comes with an AI answer. The **Citations** page shows which third-party pages AI engines trust to talk about your brand.
+
+## Prompts that skew your metrics
+
+Prompts that include your brand name almost always score high, because the AI obviously talks about you when asked. They don't measure how AI engines discover you on their own, and they distort your averages, so keep tracked prompts free of your brand name.
diff --git a/web/src/lib/help/articles.ts b/web/src/lib/help/articles.ts
new file mode 100644
index 000000000..09d1b17e7
--- /dev/null
+++ b/web/src/lib/help/articles.ts
@@ -0,0 +1,125 @@
+/**
+ * Ansvisor Help (CHAT-19 §4): one Markdown article per page, feature,
+ * metric, setting and main task, in the repo's /help/{area}/{slug}.md, that
+ * Chat answers how-to questions from (only approved ones), section by
+ * section. This reads an article's front matter and splits it into the
+ * sections retrieval returns one at a time. No dependency: the front matter
+ * is a small fixed subset of YAML (`key: value`, `key: [a, b]`, booleans).
+ */
+
+/** The folders under /help: one per area of the app. */
+export const HELP_AREAS = [
+ 'getting-started',
+ 'visibility',
+ 'prompts',
+ 'mentions',
+ 'citations',
+ 'competitors',
+ 'content',
+ 'site-audit',
+ 'action-center',
+ 'ai-traffic',
+ 'reports',
+ 'settings',
+] as const;
+export type HelpArea = (typeof HELP_AREAS)[number];
+
+/** The lowest plan an article's feature needs; `all` for every plan and self-hosted. */
+export const HELP_PLANS = ['all', 'starter', 'growth', 'enterprise'] as const;
+export type HelpPlan = (typeof HELP_PLANS)[number];
+
+/** Words a section may have at most, so retrieval can return one alone (§4). */
+export const MAX_SECTION_WORDS = 150;
+
+export interface HelpFrontMatter {
+ title: string;
+ /** The page it is about, which Chat's navigation card opens. */
+ route: string;
+ keywords: string[];
+ synonyms: string[];
+ plan: HelpPlan;
+ /** YYYY-MM-DD: when the article last changed. */
+ updated: string;
+ /** YYYY-MM-DD: when it was last checked against the app. */
+ last_checked: string;
+ owner: string;
+ /** The app version it was checked against (package.json). */
+ app_version: string;
+ /** Chat answers only from approved articles. */
+ approved: boolean;
+ /** The documentation it was written from (repo paths). */
+ sources: string[];
+}
+
+export interface HelpSection {
+ /** Null for the text before the first heading. */
+ heading: string | null;
+ text: string;
+}
+
+export interface HelpArticle {
+ area: string;
+ slug: string;
+ meta: Partial<HelpFrontMatter>;
+ sections: HelpSection[];
+}
+
+const LIST_FIELDS = new Set(['keywords', 'synonyms', 'sources']);
+
+function parseValue(key: string, raw: string): unknown {
+ const value = raw.trim();
+ if (LIST_FIELDS.has(key)) {
+ if (!value.startsWith('[') || !value.endsWith(']')) return undefined;
+ const inner = value.slice(1, -1).trim();
+ return inner === ''
+ ? []
+ : inner.split(',').map((item) => item.trim().replace(/^["']|["']$/g, ''));
+ }
+ if (value === 'true' || value === 'false') return value === 'true';
+ return value.replace(/^["']|["']$/g, '');
+}
+
+/** An article's front matter and sections; `path` is help/{area}/{slug}.md. */
+export function parseHelpArticle(path: string, text: string): HelpArticle {
+ const parts = path.replace(/\\/g, '/').split('/');
+ const slug = (parts.at(-1) ?? '').replace(/\.md$/, '');
+ const area = parts.at(-2) ?? '';
+ const meta: Record<string, unknown> = {};
+ let body = text;
+ const match = /^---\n([\s\S]*?)\n---\n?/.exec(text);
+ if (match) {
+ for (const line of match[1].split('\n')) {
+ const at = line.indexOf(':');
+ if (at <= 0) continue;
+ const key = line.slice(0, at).trim();
+ meta[key] = parseValue(key, line.slice(at + 1));
+ }
+ body = text.slice(match[0].length);
+ }
+ const sections: HelpSection[] = [];
+ let current: HelpSection = { heading: null, text: '' };
+ for (const line of body.split('\n')) {
+ const heading = /^##\s+(.+)$/.exec(line);
+ if (heading) {
+ if (current.heading !== null || current.text.trim()) sections.push(current);
+ current = { heading: heading[1].trim(), text: '' };
+ } else {
+ current.text += `${line}\n`;
+ }
+ }
+ if (current.heading !== null || current.text.trim()) sections.push(current);
+ return {
+ area,
+ slug,
+ meta: meta as Partial<HelpFrontMatter>,
+ sections: sections.map((s) => ({ ...s, text: s.text.trim() })),
+ };
+}
+
+/** Words in a section, as a reader counts them: Markdown marks aside. */
+export function wordCount(text: string): number {
+ return text
+ .replace(/[#*_>`[\]()|-]/g, ' ')
+ .split(/\s+/)
+ .filter((word) => /[\p{L}\p{N}]/u.test(word)).length;
+}
diff --git a/web/src/lib/help/validate.test.ts b/web/src/lib/help/validate.test.ts
new file mode 100644
index 000000000..7c2ded43a
--- /dev/null
+++ b/web/src/lib/help/validate.test.ts
@@ -0,0 +1,180 @@
+// @vitest-environment node
+import { existsSync, readdirSync, readFileSync, statSync } from 'node:fs';
+import path from 'node:path';
+import { describe, expect, it } from 'vitest';
+import {
+ HELP_AREAS,
+ HELP_PLANS,
+ MAX_SECTION_WORDS,
+ parseHelpArticle,
+ wordCount,
+ type HelpArticle,
+} from './articles';
+import { SETTINGS_SECTIONS } from '@/app/[locale]/(dashboard)/dashboard/settings/_sections';
+
+/**
+ * Every Help article (repo /help, CHAT-19 §4) keeps the format Chat's Help
+ * answers rely on: its front matter, a route to a page of the app, short
+ * sections, and nothing internal (CHAT-19 §3 "Never shown").
+ */
+
+const REPO = path.resolve(__dirname, '../../../..');
+const HELP = path.join(REPO, 'help');
+const APP = path.join(REPO, 'web/src/app/[locale]/(dashboard)');
+
+function articleFiles(dir = HELP): string[] {
+ if (!existsSync(dir)) return [];
+ return readdirSync(dir).flatMap((entry) => {
+ const full = path.join(dir, entry);
+ if (statSync(full).isDirectory()) return articleFiles(full);
+ return entry.endsWith('.md') && full !== path.join(HELP, 'README.md') ? [full] : [];
+ });
+}
+
+/** The app's page for a route: a folder per segment, or a [param] folder for any. */
+function pageExists(route: string): boolean {
+ let dirs = [APP];
+ for (const segment of route.split('?')[0].split('/').filter(Boolean)) {
+ dirs = dirs.flatMap((dir) => {
+ if (!existsSync(dir)) return [];
+ const named = path.join(dir, segment);
+ const params = readdirSync(dir)
+ .filter((entry) => /^\[[^.\]]+\]$/.test(entry))
+ .map((entry) => path.join(dir, entry));
+ return [...(existsSync(named) ? [named] : []), ...params];
+ });
+ }
+ return dirs.some((dir) => existsSync(path.join(dir, 'page.tsx')));
+}
+
+const DATE = /^\d{4}-\d{2}-\d{2}$/;
+/** Words that would show internals in an answer (CHAT-19 §3). */
+const INTERNAL =
+ /\b(supabase|postgres(ql)?|sql|migration|database table|api\/|localhost|organization_id|brand_id|json|webhook secret|stack trace)\b/i;
+
+function problemsOf(file: string, article: HelpArticle): string[] {
+ const at = path.relative(REPO, file);
+ const out: string[] = [];
+ const m = article.meta;
+ const say = (problem: string) => out.push(`${at}: ${problem}`);
+
+ if (!(HELP_AREAS as readonly string[]).includes(article.area))
+ say(`unknown area ${article.area}`);
+ if (!/^[a-z0-9]+(-[a-z0-9]+)*$/.test(article.slug)) say('slug is not kebab-case');
+ if (typeof m.title !== 'string' || !m.title || m.title.length > 80)
+ say('title missing or over 80');
+ if (typeof m.route !== 'string' || !m.route.startsWith('/dashboard')) {
+ say('route must be an app page under /dashboard');
+ } else {
+ if (!pageExists(m.route)) say(`route ${m.route} is no page of the app`);
+ const section = /^\/dashboard\/settings\/([^/?]+)/.exec(m.route)?.[1];
+ if (section && !(SETTINGS_SECTIONS as readonly string[]).includes(section)) {
+ say(`route ${m.route} names no Settings section`);
+ }
+ }
+ if (!Array.isArray(m.keywords) || m.keywords.length < 3) say('needs 3 keywords or more');
+ if (!Array.isArray(m.synonyms)) say('synonyms must be a list');
+ if (!(HELP_PLANS as readonly string[]).includes(m.plan as string))
+ say('plan must be all, starter, growth or enterprise');
+ for (const key of ['updated', 'last_checked'] as const) {
+ const value = m[key];
+ if (typeof value !== 'string' || !DATE.test(value) || Number.isNaN(Date.parse(value))) {
+ say(`${key} must be YYYY-MM-DD`);
+ }
+ }
+ if (typeof m.owner !== 'string' || !m.owner) say('owner missing');
+ if (typeof m.app_version !== 'string' || !/^\d+\.\d+\.\d+$/.test(m.app_version)) {
+ say('app_version must be x.y.z');
+ }
+ if (typeof m.approved !== 'boolean') say('approved must be true or false');
+ if (!Array.isArray(m.sources) || m.sources.length === 0)
+ say('sources must list what it was written from');
+ for (const source of Array.isArray(m.sources) ? m.sources : []) {
+ if (!existsSync(path.join(REPO, source))) say(`source ${source} doesn't exist`);
+ }
+
+ const headed = article.sections.filter((s) => s.heading);
+ if (headed.length < 2) say('needs two sections or more');
+ for (const section of article.sections) {
+ const name = section.heading ?? '(intro)';
+ const words = wordCount(section.text);
+ if (words === 0) say(`section ${name} is empty`);
+ if (words > MAX_SECTION_WORDS)
+ say(`section ${name} has ${words} words (max ${MAX_SECTION_WORDS})`);
+ }
+ const body = article.sections.map((s) => `${s.heading ?? ''}\n${s.text}`).join('\n');
+ if (/`/.test(body)) say('no code: backticks are not allowed');
+ if (/\]\(/.test(body)) say('no links: Chat adds the page card from route');
+ if (/<[a-z][^>]*>/i.test(body)) say('no HTML');
+ if (/\/dashboard\b/.test(body)) say('name pages by their names, not their addresses');
+ const internal = INTERNAL.exec(body);
+ if (internal) say(`internal detail "${internal[0]}"`);
+ return out;
+}
+
+describe('parseHelpArticle', () => {
+ it('reads the front matter and splits the sections, the intro first', () => {
+ const article = parseHelpArticle(
+ 'help/prompts/add-prompts.md',
+ [
+ '---',
+ 'title: Add prompts',
+ 'route: /dashboard/prompts/tracked',
+ 'keywords: [prompts, add, "bulk import"]',
+ 'synonyms: []',
+ 'approved: false',
+ '---',
+ 'Prompts are the questions Ansvisor asks AI engines.',
+ '',
+ '## One at a time',
+ 'Click Add prompt.',
+ '## Many at once',
+ 'Use Bulk import.',
+ ].join('\n'),
+ );
+ expect(article).toEqual({
+ area: 'prompts',
+ slug: 'add-prompts',
+ meta: {
+ title: 'Add prompts',
+ route: '/dashboard/prompts/tracked',
+ keywords: ['prompts', 'add', 'bulk import'],
+ synonyms: [],
+ approved: false,
+ },
+ sections: [
+ { heading: null, text: 'Prompts are the questions Ansvisor asks AI engines.' },
+ { heading: 'One at a time', text: 'Click Add prompt.' },
+ { heading: 'Many at once', text: 'Use Bulk import.' },
+ ],
+ });
+ });
+
+ it('counts the words a reader reads', () => {
+ // The dash is a mark, not a word.
+ expect(wordCount('**Add** a prompt — in 2 clicks.')).toBe(6);
+ });
+});
+
+describe('the Help articles', () => {
+ const files = articleFiles();
+ const articles = files.map((file) => ({
+ file,
+ article: parseHelpArticle(file, readFileSync(file, 'utf8')),
+ }));
+
+ it('cover the live app: forty articles or more', () => {
+ expect(articles.length).toBeGreaterThanOrEqual(40);
+ });
+
+ it('keep the format Chat answers from, and nothing internal', () => {
+ const problems = articles.flatMap(({ file, article }) => problemsOf(file, article));
+ expect(problems).toEqual([]);
+ });
+
+ it('give each article its own slug and title', () => {
+ const repeated = (values: string[]) => values.filter((value, i) => values.indexOf(value) !== i);
+ expect(repeated(articles.map(({ article }) => article.slug))).toEqual([]);
+ expect(repeated(articles.map(({ article }) => String(article.meta.title)))).toEqual([]);
+ });
+});