claude-seo-ai

The SEO + AI-search optimization toolkit for Claude Code.
Audit any website on two independent axes — classic Search SEO and AI Visibility (GEO/AEO) — and optionally apply the safe fixes for you.

Two scores, never blended. A page can rank well in Google yet be uncitable by ChatGPT, Perplexity, Google AI Overviews, Gemini, and Claude — or the reverse. claude-seo-ai measures both and tells you exactly what to fix.

🇪🇸 Resumen en español al final · guías completas en docs/es/.

Why

AI answer engines now sit beside classic search as a primary discovery channel, and the rules for being cited differ from the rules for ranking. Most SEO tools treat AI search as a footnote and never write fixes back into your code. This one is built for 2026–2027: honest, agent-native, and able to both diagnose and (opt-in) fix.

It is original work — inspired by the patterns of community SEO tooling but copies no branding, text, or names from any other project. MIT-licensed.

Install

As a Claude Code plugin (recommended):

/plugin marketplace add Hainrixz/claude-seo-ai
/plugin install claude-seo-ai@claude-seo-ai
/reload-plugins

Cross-agent (Cursor, Codex, Gemini CLI, Windsurf…) via Vercel Skills:

npx skills add Hainrixz/claude-seo-ai

The plugin works fully offline (Tier 0, no keys). See Data tiers for optional rendering and API integrations.

Usage

/claude-seo-ai:audit  <url|path> [--render auto|static|js] [--vertical auto|saas|blog-publisher|local-business|ecommerce|docs]
/claude-seo-ai:geo    <url|path>            # AI-search (GEO/AEO) subset → AI Visibility score
/claude-seo-ai:score                         # recompute/show scores from the last audit
/claude-seo-ai:fix    <url|path> [--category schema|meta|robots|sitemap|hreflang|alt|canonical|social|llms] [--dry-run]

audit is read-only and never touches your files. fix previews diffs by default and writes only after you confirm each change. Power users can call any module directly, e.g. /claude-seo-ai:seo-schema-jsonld.

Two scores, never blended

Every audit reports two 0–100 scores with letter bands (A–F) and a one-line interpretation (details):

• Search SEO — weighted toward indexability, Core Web Vitals, on-page, schema.
• AI Visibility (GEO/AEO) — weighted toward answer extractability, schema, fact density, AI-crawler access, entities.

Severity gating caps a score at F if something critical fails (e.g. site-wide noindex). Conditional verticals re-normalize so a blog isn't penalized for lacking Product schema. needs_api checks are excluded from the math and counted separately as score confidence.

How it works

A skill-first, three-layer design (Claude is the runtime; the Node helpers are optional):

1. Directive — one of the four command skills (audit/geo/score/fix).
2. Orchestration — seo-orchestrator builds a shared PageSnapshot, detects the site vertical, and dispatches read-only specialist subagents in parallel.
3. Execution — ~20 focused seo-* modules each emit findings conforming to schema/finding.schema.json — with observed evidence and a runnable verification.reproduce command. See docs/en/architecture.md.

What it audits

A complete 2026 suite, grouped:

Get cited by AI engines

The GEO/AEO modules score how citable your content is: self-contained answer blocks, fact density, original data, entity linking, and whether retrieval bots (OAI-SearchBot, Claude-SearchBot, PerplexityBot) can actually reach and render your page. See references/ai-crawlers.md.

The opt-in fixer

The fixer (skills/fix) is disable-model-invocation: true — Claude can never trigger writes on its own. Only /claude-seo-ai:fix does, and only the one writer subagent (seo-fixer-writer) has Write/Edit; every auditor is read-only by tool allowlist.

• AUTO (deterministic, additive, verifiable): meta viewport/charset/lang, Tier-1 JSON-LD, robots.txt AI directives, canonical, hreflang sets, OG/Twitter cards, image dimensions, XML sitemaps, llms.txt (disclosure-gated).
• PROPOSED (per-item accept): generated <title>/meta description, answer-block/TL;DR rewrites, internal-link insertions, heading restructuring, generated image alt text.
• ADVISORY (never written): content rewrites, Core Web Vitals/performance, rendering strategy, redirects, link-building, Merchant Center/GBP backend.

Dry-run by default · unified-diff preview · git-aware (refuses a dirty tree) · backups under ${CLAUDE_PLUGIN_DATA} · idempotent · re-verifies each change · never writes to .git/secrets/lockfiles (enforced by a PreToolUse hook).

Honesty guardrails

This tool refuses to ship SEO myths:

• llms.txt is generated but scored 0 and labeled low/uncertain impact (~10% adoption; partial vendor support — Anthropic and Perplexity honor it in retrieval, Google does not use it for AI Overviews/AI Mode, OpenAI uncommitted — also useful as IDE-agent context).
• FAQPage/HowTo flagged as deprecated-for-Google-rich-results (still valid for AI extraction) — never counted as a rich-result win.
• Keyword density is not optimized for; stuffing is flagged as a negative.
• No "AI-specific keyword" magic — citation comes from extractable structure, verifiable facts, and authority.
• Lab vs field Core Web Vitals are clearly distinguished (only CrUX field p75 drives the score).
• No fabrication, ever — never invents statistics, citations, dates (no backdating), credentials, or sameAs identity links.
• Every finding carries a confidence tier (established / directional / speculative); leaked/inferred guidance (e.g. NavBoost) ships only as directional.

Data tiers

See .mcp.json.example for opt-in render MCPs and docs/en/mcp.md.

Optional scripts

The skills work as pure Markdown; the zero-dependency Node helpers in scripts/ sharpen accuracy and make verification.reproduce runnable (Node ≥ 18, no install step):

node scripts/parse-html.mjs           --file tests/fixtures/blog-post.html
node scripts/validate-jsonld.mjs      --url  https://example.com
node scripts/parse-robots-sitemap.mjs --url  https://example.com
node scripts/score.mjs                --findings findings.json
node tests/run.mjs                    # 30 assertions over the fixtures

Project structure

.claude-plugin/   plugin.json + marketplace.json
skills/           4 command skills (audit, geo, score, fix → /claude-seo-ai:audit …)
                  + orchestrator, vertical-detect, crawl-render, seo-score + ~20 audit modules
agents/           5 subagents (4 read-only auditors + 1 writer)
hooks/            PreToolUse write guard
scripts/          optional zero-dep Node helpers (+ verification)
references/       scoring model, AI crawlers, schema, CWV, routing
schema/           finding + report JSON Schemas, JSON-LD templates
docs/en, docs/es  bilingual guides
assets/           pixel-art images
tests/fixtures    sample sites for verification

About

Built by Enrique Rocha — I help teams ship AI: consulting, automations, and agents. This is a community, MIT-licensed project: use it, fork it, open issues and PRs (see CONTRIBUTING.md).

• 🌐 tododeia.com
• 📸 Instagram @soyenriquerocha

License

MIT · Claude mascot artwork generated for this project in pixel-art style.

🇪🇸 Resumen (Español)

claude-seo-ai es la herramienta open-source de SEO + búsqueda con IA para Claude Code. Audita cualquier sitio en dos puntajes independientes — Search SEO clásico y Visibilidad en IA (GEO/AEO) — con hallazgos reproducibles, y opcionalmente aplica las correcciones seguras por ti (meta tags, JSON-LD, robots.txt para crawlers de IA, hreflang, sitemaps…), siempre con confirmación previa.

• Instalar: /plugin marketplace add Hainrixz/claude-seo-ai → /plugin install claude-seo-ai@claude-seo-ai. Cross-agente: npx skills add Hainrixz/claude-seo-ai.
• Usar: /claude-seo-ai:audit <url> · :geo · :score · :fix (vista previa por defecto; nunca escribe sin tu confirmación).
• Honestidad: llms.txt se genera pero puntúa 0; FAQ/HowTo marcados como sin rich results; nunca inventa estadísticas, fechas ni enlaces de identidad.
• Sin claves funciona (Tier 0, offline). Render de SPAs y Core Web Vitals reales son opcionales (Tier 1+).

Guías completas en español: docs/es/ · hecho con cariño por tododeia.com.