Grade any website's AI readiness — 24 checks across crawlability, content quality, citation signals, and AI metadata. Every check cites its source. No opinions.
claude /install ai-grader@andorlabsOr load for a single session:
claude --plugin-dir /path/to/ai-grader/ai-grader:grade <url>
One command does everything: tests AI bot access, scores all 24 checks, prints a prioritized scorecard to the terminal, and writes a full detailed report to ~/Desktop/claude-code/.
| Category | Checks | What It Evaluates |
|---|---|---|
| Crawlability Signals | 7 | WAF/CDN bot access, robots.txt AI policy, sitemap, meta robots, SSR rendering, JSON-LD schema, OG metadata alignment |
| Content Quality for LLMs | 6 | Semantic HTML, heading hierarchy, answer-first content, FAQ sections, content-to-chrome ratio, entity consistency |
| Citation Readiness | 6 | External citations, author attribution, content freshness, original research, claim statements, brand disambiguation |
| AI-Specific Metadata | 5 | llms.txt, AI usage policy, API documentation, MCP/tool-use readiness, content licensing |
Every check is grounded in empirical evidence:
- Princeton/Georgia Tech GEO Study (KDD 2024) — the only peer-reviewed generative engine optimization research
- Ahrefs 75,000-Brand Study (2025) — brand mentions vs. AI visibility correlation
- SE Ranking Study (~300K domains) — content freshness and AI citation patterns
- Victorino Group Analysis (1.2M ChatGPT responses) — entity density and citation language
- Official documentation from OpenAI, Anthropic, Google, Perplexity, Bing, schema.org
Every check produces:
- What it is — from the reference (word for word)
- Why it matters — the rationale with research citations
- Source — full citation (org, document, year)
- Compliance — PASS / FAIL / N/A
- Observation — what was found on the site
- How to fix — actionable recommendation (FAIL only)
- Claude Code CLI
- WebFetch (built-in)
- Bash (built-in, for AI bot user-agent testing)
- Chrome (optional, for homepage screenshot)
MIT