The 12-point GEO audit: can AI engines actually cite you?
Before you write another word of content, find out whether AI engines can even find, parse, and trust your pages. These are the twelve checks we run on every site — grouped into access, content, structure, and authority — with the fix for each.
A GEO audit answers one question: can AI engines find, parse, trust, and cite your pages? Run twelve checks across four layers — access (crawlable, indexed, server-rendered), content (self-contained answers, verifiable claims), structure (headings, lists, valid schema), and authority (entities, llms.txt, corroboration) — and fix failures top-down.
A GEO audit checks whether AI engines can find, parse, trust, and cite your pages. It shares SEO's technical base but adds the checks that decide whether you become the answer, not just a ranked link. Work through all twelve in order — a failure early on (say, you're not crawlable) makes everything below it moot.
What are the key components of a GEO audit for a client?
A GEO audit for a client has four key components, checked in this order: access (AI crawlers allowed, page indexed, answer in server-rendered HTML), content (self-contained answers, verifiable claims, question-style headings), structure (lists and tables, valid JSON-LD, clean internal links), and authority (defined entities, llms.txt, corroboration across the web).
Those four components hold the twelve checks below. The same list works for an agency auditing a client, a consultant, or an in-house team auditing its own site.
- Access (checks 1–3): Can AI engines reach and read the page?
- Content (checks 4–6): Is there a clear answer an engine can lift?
- Structure (checks 7–9): Can the answer be extracted cleanly?
- Authority (checks 10–12): Can the source be trusted?
What should the client deliverable contain?
A client GEO audit deliverable should give every check four things, so the client can act on it without a follow-up call:
- Findings: pass, partial, or fail for each of the twelve checks, with the page and the evidence you saw.
- Priority order: failures sorted top-down, access first and authority last, because an early failure blocks everything below it.
- Owner: who fixes each item: a developer for crawling and rendering, a writer for answers and headings, and whoever runs profiles and PR for authority.
- How to re-check: the exact test to run again after the fix, and the citation share baseline to compare against.
Access — can engines reach you?
Engines can reach you when their crawlers are allowed, the page is indexed, and the answer is in the HTML they fetch.
1. AI crawlers are allowed
Check: your robots.txt lets in the search crawlers behind AI answers: OAI-SearchBot for ChatGPT search (OpenAI), Claude-SearchBot (Anthropic), PerplexityBot (Perplexity), and Googlebot, which feeds AI Overviews and AI Mode. Fix: explicitly allow the AI user-agents you want citing you — if they're blocked, nothing else matters (it's the first reason a brand isn't in ChatGPT). Training crawlers are a separate choice: blocking GPTBot or ClaudeBot opts you out of model training, not search answers. Google-Extended covers Gemini training and grounding, and Google says it does not affect inclusion in Google Search (Google).
2. The page is indexed and fast
Check: the URL is indexed, loads quickly, and has clean canonicals. Fix: resolve crawl errors, duplicate canonicals, and slow render — AI Overviews retrieve from the existing index, and Google says a page must be indexed and eligible to show with a snippet (Google). Also confirm your site is still included in Search Console's Search generative AI control, which reached every site on August 31, 2026.
3. The answer is in server-rendered HTML
Check: view source (not the inspector) and confirm the actual answer text is present. Fix: if content is injected by client-side JavaScript an engine may not run, pre-render or server-render it.
Content — is there an answer to lift?
There's an answer to lift when each section opens with a clear, self-contained, verifiable statement that matches how people ask.
4. Each section leads with a self-contained answer
Check: the first 40–60 words of a section answer the question on their own. Fix: move the answer above the backstory; lead, then elaborate.
5. Claims are unambiguous and verifiable
Check: no hedging or vague superlatives engines won't repeat. Fix: make one clear, defensible statement per question — if it can't be defended, it won't be cited.
6. Headings match how people ask
Check: question-phrased H2/H3s mirror real queries — the core of answer engine optimization. Fix: rename "Our approach" to the actual question a buyer types.
Structure — can it be extracted?
It can be extracted when lists, tables, schema, and page architecture make each answer easy to find and copy.
7. Lists and tables for enumerable content
Check: comparisons and steps use real <table>/<ol> markup. Fix: convert div-grids and prose lists into semantic lists and tables engines can lift verbatim.
8. Valid JSON-LD that matches the page
Check: Article/FAQ/HowTo schema present and identical to the visible text. Fix: add the right type and keep schema and copy in exact agreement. Google's guidelines say “Don't mark up content that is not visible to readers of the page.” (Google)
9. Clean architecture and internal links
Check: one clear topic per page, linked to related pages. Fix: consolidate thin/duplicate pages and link the cluster so engines map your topical authority.
Entities & authority — can it be trusted?
It can be trusted when engines can tell exactly who you are and other credible sources say the same things about you.
10. A defined Organization and entities
Check: Organization schema with sameAs to your real profiles, plus clear entities for what you do. Fix: disambiguate who you are so engines trust the source, not just the string of text.
11. llms.txt and agents.md exist
Check: machine-readable companions are published and linked. Fix: add an llms.txt brand summary and agents.md so AI agents can read a clean version of who you are. Keep expectations honest: llms.txt is still “a proposal” (llmstxt.org), and Google says Google Search ignores it, so it neither helps nor harms your AI Overview odds (Google). Keep it anyway: Google says it's "fine" to maintain these files "for other services or systems that use them" (Google), and AI agents do read them. For Google, it's neutral; for everyone else, it's reach. That split runs through this whole list; see how to get cited by every AI engine.
12. Corroboration across the web
Check: credible links, mentions, and a consistent brand entity. Fix: earn citations and keep your name, claims, and profiles consistent so other trusted sources agree with you.
How do you score your GEO audit?
Mark each check pass / partial / fail, then fix failures top-down — access before content, content before structure, structure before authority. There's no point polishing schema on a page an engine can't crawl. Re-test after each change and watch citation share move — here's how to measure AI visibility across engines. For Google, the Generative AI performance report in Search Console shows how often your pages appeared in AI Overviews and AI Mode.
Most brands fail the same check first: there's no clean answer on the page for an engine to lift. Fix that one thing and everything downstream gets easier.
Sources
- Overview of OpenAI Crawlers — OpenAI. Read Sep 30, 2026.
- Does Anthropic crawl data from the web, and how can site owners block the crawler? — Anthropic. Read Sep 30, 2026.
- Perplexity Crawlers — Perplexity. Read Sep 30, 2026.
- List of Google's common crawlers — Google for Developers. Read Sep 30, 2026.
- Optimizing your website for generative AI features on Google Search — Google Search Central. Read Sep 30, 2026.
- Search generative AI control — Search Console Help. Read Sep 30, 2026.
- General structured data guidelines — Google Search Central. Read Sep 30, 2026.
- The /llms.txt file, v2 — llmstxt.org. Read Sep 30, 2026.
- Latest Google Search Documentation Updates — Google Search Central. Read Sep 30, 2026.
Key takeaways
- A GEO audit asks four questions in order: can engines reach, read, extract, and trust your pages?
- Fix failures top-down — crawl access first, authority last; early failures void everything below.
- The most common failure is a buried answer — lead each section with a direct 40–60 word response.
- Re-audit on every content change and quarterly site-wide; track citation share continuously.
Questions about GEO audits.
What is a GEO audit?
A GEO (Generative Engine Optimization) audit checks whether AI engines can find, parse, trust, and cite your site. It reviews crawl access for AI bots, whether answers are stated in clean server-rendered HTML, how content is structured for extraction, entity and structured-data clarity, and the authority signals that make a page citable across ChatGPT, Google AI Overviews, Perplexity, and others.
How is a GEO audit different from an SEO audit?
An SEO audit focuses on ranking in the list of results — keywords, links, Core Web Vitals, indexation. A GEO audit shares that technical base but adds checks specific to being the cited answer: AI-bot access, self-contained passages, entity clarity, llms.txt, and citation share. SEO is the foundation; the GEO audit measures extractability and trust on top of it.
Can I run a GEO audit myself?
Yes. Work through the 12 checks above in order — access, content, structure, entities/authority — and note pass/fail with the fix for each. A free analyzer can flag the technical items; the judgment calls (is the answer truly self-contained? is the claim verifiable?) are best reviewed by a person.
How often should I run a GEO audit?
Re-audit any page when you change its content or structure, and run a full site pass quarterly — AI engines, their crawlers, and their answer formats change quickly. Track citation share continuously so you catch drops between audits.
What's the most common GEO audit failure?
Burying the answer. Most pages open with brand backstory or a long intro, so there is no clean, self-contained passage for an engine to lift — even when the page ranks. The single highest-leverage fix is to lead each section with a direct 40–60 word answer.
What should a GEO site audit checklist for AI search optimization include?
A GEO site audit checklist for AI search optimization should include twelve checks in four groups: access (AI crawlers allowed, indexed, server-rendered), content (self-contained answers, verifiable claims, question headings), structure (lists and tables, valid JSON-LD, internal links), and authority (entities, llms.txt, corroboration). Score each pass, partial, or fail, then fix failures top-down.
What should a client GEO audit report include?
A client GEO audit report should include the findings for every check (pass, partial, or fail, with evidence), the fixes in priority order, an owner for each fix, and how to re-check it, including the citation share baseline. That turns the audit into a work plan the client can hand straight to their team.
Can GEO platforms run the whole audit automatically?
No. A GEO platform can automate the technical checks, such as crawler access, indexing, rendering, schema validity, and llms.txt, and it can track citation share over time. The judgment calls still need a person: whether an answer is truly self-contained, and whether a claim is clear and defensible.
Related posts & guides.
Get the audit run for you — across every AI engine.
We'll work through all twelve checks on your site, show you exactly where AI engines can't cite you, and hand you a prioritized fix list within 24 hours.