Skip to content

armgit / macro

Public Imported from GitHub
0

Macro is a unified workspace for teams: email, chat, docs, tasks, agents, calls, and CRM — @-linked together with shared AI memory.

6.8 KB
Raw Blame

Public-site SEO / AEO migration audit

Audited September 23, 2026 against the live https://macro.com/sitemap.xml, its 30 listed pages, robots.txt, the imported marketing source, and the previous solid-site CloudFront configuration. No production infrastructure was changed.

See the September 24 recheck for the live crawl and the September 25 standalone verification for current architecture, artifact sizes, Lighthouse measurements, and deployment checks.

Preserved and repaired

  • All 30 indexed URLs remain at their existing paths. No feature/blog URL needs a migration redirect. This includes /agents, /github, /migrate, pricing, partner/legal pages, and all 11 posts, including the five alternative pages.
  • The feature/blog prerenderer, titles, descriptions, canonicals, social images, article metadata, breadcrumbs, SoftwareApplication and FAQ markup remain. Build checks compare titles and schema types with the captured live baseline.
  • Fixed the new homepage's empty HTML shell and missing Organization, WebSite and SoftwareApplication data. Its actual public UI is now prerendered for everyone; visitors and non-JavaScript crawlers receive the same content.
  • Navigation links now exist in the initial HTML even while the menu is closed. Feature and comparison pages retain their original contextual links and FAQs. Restored the missing Startups menu link; the build also follows the HTML link graph and requires all 30 baseline pages to be reachable from the homepage.
  • /start and /start.html now have their own canonical and noindex, follow. They no longer masquerade as the homepage. Existing mobile-signup noindex and sitemap exclusions remain.
  • Production robots.txt still allows public crawling, excludes /app and /test, and advertises the sitemap. Non-production builds still disallow all. No new bot-specific restrictions were added. The live site returned 403 for /llms.txt, /llms-full.txt and /agents.md; there were no working files at those URLs to migrate, and none exist in the old public assets directory.
  • Restored the marketing analytics initialization and Meta Pixel on the public journey. The existing GA4, Ads and PostHog configuration is reused.
  • Nearby demo code preloads sequentially during idle time after page load (except on Save-Data/2G connections); editors only mount near the viewport, with section copy always present and quiet placeholders while loading. Demos, components, styles, and assets are website-owned, with no application source dependencies. Initial JS is 192 KB gzip after the September 25 isolation, versus more than 1.9 MB before the first pass; the build now fails above 225 KB. The homepage hydrates its server-rendered content, prioritizes CSS/fonts, and loads PostHog after paint while preserving queued events and attribution. These artifact sizes and local lab results do not establish production Core Web Vitals.

Hosting contract — required before switching traffic

The current monorepo app deploy does not publish the public site. Keep the old site origin available until the candidate passes the checks below.

  1. Install Chromium (bunx playwright install chromium --only-shell --with-deps on Linux CI), fetch LFS assets, build with VITE_APP_BASE_URL=https://macro.com and run bun run build in apps/marketing. Publish apps/marketing/dist/ to the website origin; keep the app's apps/web/dist/ on the existing app origin.
  2. Retain direct 200 directory-index rewrites: /email → /email/index.html, /posts/linear-alternative → /posts/linear-alternative/index.html, etc. Do not send all public URLs to the homepage SPA fallback. Unknown URLs must return a real 404/410, not a homepage with status 200.
  3. Preserve the existing special behaviors: /app, /app/*, app static files, /.well-known/* OIDC, and resources* to the Ghost origin. The legacy /resources content is outside the sitemap and this static build; dropping that proxy would lose pages despite the sitemap checks passing.
  4. Preserve HTTPS, www → apex with the path/query intact, chat.* → app chat, and the authenticated-root → /app redirect. These are CDN behaviors, not client-side redirects. If URLs ever change, add explicit permanent redirects to relevant replacements rather than redirecting everything to /.
  5. Keep Brotli/gzip delivery, immutable caching for hashed assets, revalidation for HTML, correct content types, and cache invalidation. Do not set a public X-Robots-Tag: noindex. Keep Search Console/Bing DNS verification and any CDN verification responses. No verification HTML/meta files were found in the old source; ownership settings cannot be verified from this repository.
  6. Retain crawler access through WAF/CDN, not just robots.txt. Confirm Googlebot, Bingbot and relevant AI search crawlers using the providers' verification methods; spoofing a User-Agent alone does not establish real bot access.

Cutover verification

  • Run bun run check:seo:live https://candidate-host against the candidate CDN, then again against https://macro.com immediately after cutover. It checks all 30 paths without JavaScript/cookies, metadata, sitemap, robots, /start, and unknown-path status. The local filesystem gate cannot validate CDN rules.
  • Separately exercise the existing /resources/* Ghost pages, app deep links, auth redirects, www/HTTPS redirects and campaign-query preservation.
  • Run a mobile Lighthouse/PageSpeed comparison on the deployed old and candidate sites with the same settings; check LCP, INP, CLS and actual transfer/cache headers. Test a cold visit, menu links, native scrolling, editor loading and the footer CTA. The interactive app still needs JS; primary page copy does not.
  • In Search Console/Bing Webmaster Tools, inspect /, a feature page and a post, verify the canonical/rendered HTML, and resubmit the unchanged sitemap. Save the previous 28 days of queries, landing-page clicks/impressions and conversions. Compare after rollout and watch indexing, 404s, crawler errors and web vitals.
  • Keep the previous deployment for rollback. This pass prevents known technical regressions; it cannot guarantee unchanged rankings/citations after a redesign.

For AEO, the important preserved assets are fetchable text, clear feature pages, comparison content, FAQs, internal links and accurate structured data. Google does not require special AI files or extra schema for AI Overviews/AI Mode: AI features guidance. See also JavaScript SEO and site migration guidance.