Grade Scale
A90–100Agent-ReadyFully optimised for AI discovery, citation, and interaction.
B75–89GoodStrong fundamentals. Minor gaps in agent integration or schema.
C60–74Needs WorkKey AI bots blocked or missing. Significant render gap.
D45–59PoorBlocking most AI traffic. No agent integration signals.
F<45InvisibleEffectively invisible to AI systems and LLM citations.
🔍DiscoverabilityCan AI crawlers find and access your site? Foundation layer — without this, no other checks matter.
39 raw pts robots.txt present (critical)robots.txt valid syntaxsitemap.xml reachable (includes sitemap-index expansion)Sitemap URL count ≥10Homepage response headers (ETag, Last-Modified)HTTP→HTTPS enforced
🤖Bot Access ControlWhich AI bots can crawl your content? Checks 12 canonical AI User-Agents against your robots.txt, inheriting User-agent: * fallbacks per RFC 9309. Highest-weight category because access is binary.
75 raw pts GPTBot (OpenAI training)OAI-SearchBot (ChatGPT browse)ChatGPT-User (on-demand)ClaudeBot (Anthropic training)Claude-SearchBot + Claude-User (browse)PerplexityBotGoogle-Extended (Gemini training)Applebot-ExtendedCCBot (Common Crawl)Meta-ExternalAgent (Llama)Bytespider (ByteDance)Training/search bot policy alignment
📡AI DiscoveryHave you published the machine-readable files LLMs use to understand your site's content scope?
20 raw pts /llms.txt published (critical — primary GEO signal, 10 pts)/llms.txt format valid (# heading + markdown links, 5 pts)/llms-full.txt published (single-file complete docs dump, 5 pts)
👁AI VisibilityCan AI parse and cite your content correctly? Structured data, entity linking quality, schema type coverage, content legibility, and emerging agent-mode signals.
~55 raw pts JSON-LD structured data quality — all schema types scored (8 pts)Entity linking quality — @id + Wikipedia/Wikidata sameAs in JSON-LD (6 pts)Open Graph meta tags (og:title, og:description, og:image, 5 pts)sameAs entity linking — meta-tag level (5 pts)Content efficiency — text/HTML ratio (JS-heavy penalty, 5 pts)FAQPage schema quality — 3.2× LLM citation multiplier (5 pts)Article/BlogPosting schema quality — author + datePublished (5 pts)Content-Signal directive in robots.txt (3 pts)WebSite schema with name + url (3 pts)BreadcrumbList schema (3 pts)VideoObject schema on video pages (3 pts)Twitter Card meta tags (2 pts)Accept: text/markdown content negotiation (2 pts)Schema.org speakable markup (2 pts)/pricing.md machine-readable pricing (2 pts)?mode=agent returns structured content (2 pts)
⚡Agent Integration (bonus)Future-ready signals for autonomous agents. Low-weight because adoption is <1% as of May 2026. We track these to reward pioneers without punishing the rest.
13 raw pts /.well-known/agent-card.json (A2A v0.3, 2 pts)/.well-known/oauth-protected-resource (RFC 9728 — MCP auth, 2 pts)OpenAPI spec at well-known path (3 pts)/.well-known/api catalog (1 pt)/.well-known/http-message-signatures-directory (Web Bot Auth, IETF draft, 1 pt)/.well-known/agent-skills/index.json (1 pt)/.well-known/mcp.json (1 pt)Listed in Smithery MCP registry (2 pts)
🛒Commerce (bonus)Machine-payable endpoint signaling + Google Merchant Center feed readiness. GMC checks only fire on domains with Product schema detected.
9 raw pts (e-commerce only) Link: rel="payment" header or Payment-Pointer header (x402, 2 pts)Google Merchant Center Atom feed link detected on product pages (3 pts)GMC required attributes coverage — id, title, price, availability, image, brand, gtin/mpn (4 pts)
🏷Product Pages (e-commerce)For domains with product pages, we auto-discover up to 10 URLs from the sitemap and audit each for Schema.org product schema completeness. Per-page checks plus an aggregate coverage score.
variable Product / ProductGroup / IndividualProduct schema present (10 pts per page)offers with price + priceCurrency + availability (8 pts)aggregateRating present (6 pts)image URL present (4 pts)brand present (3 pts)product_schema_coverage aggregate (8 pts total)
🌐GEO PresenceIs your brand visible to AI systems — in training data, in search, in category rankings? Checks cold recall via CF Workers AI, category share of voice, Wikipedia presence, brand search discoverability, plus third-party corpus signals.
46 raw pts LLM cold recall — does CF Workers AI describe the brand accurately from training data alone? (15 pts)Category share of voice — does LLM name you when asked about your category? (10 pts)Wikipedia / Wikidata entity presence — strongest predictor of LLM citation accuracy (8 pts)Hacker News mentions via hn.algolia.com — proxy for community/corpus presence (5 pts)Brand-name search returns own domain via DuckDuckGo instant (5 pts)Reddit mentions via reddit.com/search — community signal (3 pts)
Scoring model: Each check has a fixed point value (1–10 pts) based on signal importance, not pass/fail outcome. Raw scores are summed across all checks and normalized to a 100-point scale. Bonus categories (Agent Integration, Commerce) track frontier specs with <1% adoption — kept low-weight so pioneers are rewarded without penalizing the rest of the web. Bot rules inherit
User-agent: * fallbacks per RFC 9309; sitemap-index files are recursively expanded to count real URLs. Methodology version: 1.2 (May 2026).