> For the complete documentation index, see [llms.txt](https://suigar.gitbook.io/suigar-docs/llms.txt). Markdown versions of documentation pages are available by appending `.md` to page URLs; this page is available as [Markdown](https://suigar.gitbook.io/suigar-docs/seo/automation-roadmap.md). # Automation roadmap Last reviewed: 2026-05-18 This roadmap describes how Suigar moves from manual SEO docs to an AI-backed system that can plan, generate, audit, monitor, and improve SEO work without losing quality control. ## Phase 0: Memory Status: active. * Keep global rules in `guidelines.md`. * Keep channel definitions in `channels.md`. * Keep current maturity and evidence in `status.md`. * Keep project overlays in `subprojects/`. * Keep blog generation rules in `blog-generation.md`. * Keep decisions and validation in `optimization-log.md`. Exit criteria: * Every SEO task starts from this directory. * New constraints are added when learned. ## Phase 1: Structured Audits Status: next. * Add route/post audit templates that can be filled by humans or AI. * Add rendered metadata snapshots for high-value frontend and blog URLs. * Extend duplicate title/description checks to blog and docs. * Add internal-link inventory for product, docs, and blog surfaces. * Add blog post preflight checks for title, description, slug, excerpt, tags, internal links, and source ledger. Exit criteria: * Common SEO regressions are caught before merge or publish. ## Phase 2: CI And Preflight Gates Status: active. * Gate frontend route classification with existing `check:seo`. * Gate frontend title/description uniqueness and practical length bounds with `frontend/scripts/seo-metadata-report.mjs`. * Validate generated frontend static route HTML during builds with `frontend/scripts/seo-static-html.mjs`. * Deploy frontend, blog, and docs through static-site cache policy that treats HTML and SEO control files separately from immutable assets. * Run post-deploy frontend SEO checks for live route heads, robots policy, canonical host redirects, and legacy-route redirects. * Add blog metadata and sitemap/canonical consistency checks. * Add markdown/link checks for docs SEO files. * Add schema validation where structured data exists. * Add noindex/robots guardrails for testnet, previews, admin, and experimental deployments. Exit criteria: * Level 3 maturity exists for technical, crawl/indexing, canonical, metadata, and blog generation basics. ## Phase 3: Measurement Ingestion Status: active. * Import Search Console page/query data with `pnpm run seo:gsc`. * Track indexed pages, excluded pages, canonical selection, crawl issues, and query movement. * Add Core Web Vitals p75 mobile/desktop cohorts for frontend and blog. * Add GA/event data for blog engagement and product click-through. * Store enough summaries to guide AI without leaking private or irrelevant data. Exit criteria: * SEO status can cite measurement windows, not just code inspection. ## Phase 4: AI-Assisted Optimization Status: planned. * Generate content refresh candidates from Search Console and analytics. * Propose internal links from orphan pages, related docs, and topic clusters. * Propose metadata rewrites for weak click-through pages. * Propose schema additions only when visible content supports them. * Generate blog briefs from product priorities and measured content gaps. * Run AI critic passes automatically and save review artifacts. Exit criteria: * AI proposes changes from evidence; humans approve high-risk changes. ## Phase 5: Self-Improving SEO System Status: target. * AI monitors channel status, detects stale docs, and proposes constraint updates. * AI compares published content performance against intended audience/search intent. * AI updates topic clusters and blog generation rules from measured outcomes. * AI opens small, reviewable PRs for low-risk SEO maintenance. * Humans retain approval over compliance-sensitive claims, public positioning, and production publication. Exit criteria: * At least one channel reaches maturity level 5 with evidence-backed AI recommendations and human-reviewed improvements. ## Automation Backlog | Item | Channel | Risk | First Output | | ---------------------------------------------- | ------------------------------------ | ------ | --------------------------------------------------------------------- | | Rendered head audit for frontend public routes | metadata, structured-data, canonical | Medium | HTML/head snapshot report. | | Blog metadata preflight | blog-generation, metadata | Low | Sanity/local post field report. | | Blog sitemap/canonical check | crawl-indexing, canonical | Low | Build-time consistency report. | | Internal-link graph | internal-links | Low | Orphan/high-value page report. | | Social card validator | social-preview, image-seo | Medium | Preview image/copy report. | | Search Console importer | measurement | Medium | `scripts/seo/search-console-report.mjs` page/query/inspection report. | | Core Web Vitals cohort report | core-web-vitals | Medium | Mobile/desktop p75 table. | | AI content critic runner | blog-generation, content | Medium | Critique artifact per post. | | Constraint drift detector | automation, content | Medium | Suggested updates to `blog-generation.md`. | ## Guardrails * Automation must make small, reviewable changes. * Automation must never invent product facts, rewards, odds, partnerships, legal claims, or search results. * Automation must never publish generated content directly to production without the chosen review path. * Automation must record evidence and update `optimization-log.md` when it changes SEO policy or status.