The AI-Readiness Rubric (v1)

Rubric v1 — current as of 2026-07-18. Changes get noted here. 2026-08-06: the extractability denominator now excludes inline data: URI payloads (calibration fix — base64-embedded fonts/images no longer read as unextractable markup); thresholds unchanged.

100 points across 9 mechanical checks. Nothing here guarantees citations — these are the preconditions: if an answer engine can't access, parse, or extract your content, it cannot cite you, full stop. Every check is deterministic; run the audit twice and you'll get the same number.

Where the 100 points sit

llms.txt
20 pts
AI answer-engine access
15 pts
Structured data
15 pts
Title + meta description
10 pts
Heading structure
10 pts
Canonical URL
5 pts
XML sitemap
10 pts
Content extractability
10 pts
Open Graph
5 pts

What earns them

CheckPointsWhat earns them
llms.txt2015 for existing at /llms.txt (and not being a soft-404); +2.5 for starting with an H1 title; +2.5 for containing markdown links.
AI answer-engine access1515 if robots.txt exists and none of the 15 major AI crawlers are blanket-blocked; 7 if no robots.txt (defaults to allowed, but no explicit policy); 6 if some are blocked (may be intentional — the result names which bots).
Structured data15Any parseable JSON-LD on the homepage. We list the @types found.
Title + meta description105 for a 10–65-char title; 5 for a 50–165-char description; half-credit if present but out of range.
Heading structure10Exactly one H1 plus at least one H2 = 10; any H1 = 5.
Canonical URL5A rel=canonical declaration.
XML sitemap10/sitemap.xml exists and lists at least one URL.
Content extractability10Visible text ≥10% of homepage HTML = 10; ≥4% = 5; below = 0 (JS-only content is invisible to most AI crawlers). Inline data: URI payloads are excluded from the HTML denominator, so embedded fonts/images don't read as bloat.
Open Graph5Prorated across og:title, og:description, og:image.

Grade bands

A ≥85 · B 70–84 · C 55–69 · D 40–54 · F <40.

The AI crawlers we check

GPTBot, OAI-SearchBot, ChatGPT-User, ClaudeBot, Claude-SearchBot, Claude-User, PerplexityBot, Perplexity-User, Google-Extended, Applebot-Extended, CCBot, Amazonbot, DuckAssistBot, meta-externalagent, Bytespider. Blocking training bots while allowing search/user-triggered bots is a coherent posture — the audit distinguishes access status per group in the full report.

Known limits (v1)

  • Homepage-only parse (the full report re-runs checks across your top pages).
  • No JavaScript rendering — deliberately: most AI crawlers don't render either. That IS the extractability test.
  • robots.txt evaluation is prefix-heuristic (blanket Disallow: / detection), not a full RFC 9309 matcher.

Who runs this

This audit is independent and solo-operated, published by AnswerFootprint — not an SEO agency upsell, not affiliated with any AI provider. The rubric above is the whole methodology: every check is mechanical, the point weights are printed, and the known limits stay published next to the scores they qualify. Verdict looks wrong? Report it — a reproducible false pass/fail gets fixed within days and noted here.

Run the audit →