Research Report · 2026 · DeepSEOAnalysis
Benchmark data on how accessible audited websites are to AI answer engines — covering llms.txt adoption, AI crawler access, structured data coverage, and content structure. Live aggregate data from completed DeepSEOAnalysis audits is at /research.
Context
A growing share of informational queries now resolve in AI chat interfaces — ChatGPT, Perplexity, Claude, Gemini — rather than in traditional search result pages. These systems don't rank pages. They retrieve content, synthesize an answer, and cite sources. Being cited requires being accessible and extractable, not just indexable.
The majority of sites audited by DeepSEOAnalysis have at least one critical AI-visibility gap: blocked AI crawlers, no llms.txt, no FAQPage schema on pages with FAQ sections, or headings written as topics rather than questions. These are fixable — but most site owners haven't done it yet.
Methodology
These are the signals the DeepSEOAnalysis engine checks on every audit. Each has a clear pass/fail definition. The GEO score combines all five.
AI crawler access
Pass: No Disallow rules block GPTBot, ClaudeBot, PerplexityBot, or Google-Extended in robots.txt
Fail: At least one major AI crawler is explicitly blocked or blocked by a catch-all Disallow: / rule
A blocked AI crawler cannot read the page. Any content on that page is invisible to that AI system — regardless of how well-structured or authoritative it is. AI crawler access is the prerequisite for everything else.
llms.txt
Pass: yourdomain.com/llms.txt exists, returns 200, and has non-empty content
Fail: No llms.txt file found, or file returns a non-200 status
llms.txt gives AI systems a curated map of your site's most useful pages. It's the difference between an AI system guessing what your site is about and being told directly.
FAQPage / HowTo schema
Pass: At least one FAQPage or HowTo JSON-LD block is present and validates correctly
Fail: No FAQPage or HowTo schema found on any crawled page, or schema is invalid
FAQPage schema explicitly marks question-answer pairs as machine-readable. AI systems retrieving answers to questions actively look for pages with this schema — it is the strongest single signal for AI-citation readiness.
Question-form heading ratio ≥20%
Pass: On pages with 3+ H2/H3 headings: at least 20% of those headings are question-form (start with How/What/Why/Which/When/Can/Does/Is or end with ?)
Fail: Fewer than 20% of H2/H3 headings on pages with 3+ headings are question-form
AI systems retrieving answers match query intent against page structure. Headings written as questions ("How does X work?") are directly indexed against question queries. Topic-only headings ("About X") require the AI to infer the answerable question.
Content chunkability
Pass: Average word count between H2/H3 headings is under 400 words across audited pages
Fail: Average section length exceeds 400 words — long sections that mix multiple topics are harder for AI to extract from cleanly
AI retrieval systems extract fragments, not full pages. A 1,200-word section between two headings forces the AI to decide which part of it is the relevant answer — diluting precision. Shorter sections under one clear heading make extraction reliable.
Live data
The DeepSEOAnalysis research page shows live aggregate data from completed audits: llms.txt adoption rate, GPTBot and ClaudeBot access rates, average AI-visibility score, top failing checks, and CMS breakdown. The snapshot updates automatically as new audits complete.
What to do
FAQ
AI visibility measures how accessible and extractable a site's content is for AI answer engines like ChatGPT, Perplexity, and Claude. It covers five signals: AI crawler access in robots.txt, llms.txt presence, FAQPage or HowTo schema, question-form heading ratio, and content chunkability (average section length under 400 words).
The DeepSEOAnalysis GEO score checks all five AI-visibility signals and combines them into a 0–100 score. The score is reported separately from the main SEO score (which covers technical, on-page, performance, structured data, and links). See the full methodology at /methodology/ai-visibility.
AI visibility is a newer discipline than traditional SEO. Most sites were built before AI answer engines were significant traffic sources, so they lack the signals that make content extractable: no llms.txt, no FAQPage schema on pages with FAQ sections, and headings written as topics ("About Our Service") rather than questions ("What Does Our Service Do?").
FAQPage schema on pages with existing FAQ sections is typically the highest-impact change — it makes question-answer pairs machine-readable without changing any visible content. It is also the easiest to implement: generate the JSON-LD with the schema generator tool and paste it into your page head.