# AEO Audit Checklist: 34 Checks for AI Search Visibility [2026]

> A working AEO audit checklist grouped by severity — crawler access, entity resolution, extractability, freshness and attribution — with the threshold for every check.

## How to use this checklist

- Work top to bottom. The groups are ordered by severity, not by topic.

- Stop at the first blocking failure and fix it before continuing — the rest is unmeasurable until you do.

- Skip schema that does not apply. Product schema on a news article is not a gap.

- Re-check after any robots.txt, template or CMS change.

## What should an AEO audit checklist include?

An AEO audit checklist should cover five things in order: whether AI crawlers can reach the page, whether the page renders without JavaScript, whether the engine can resolve what the page is about, whether a model can extract a discrete answer from it, and whether it will attribute that answer to you. Everything below fits into one of those.

## Blocking — fix these first

Any failure here makes the rest of the checklist irrelevant.

- GPTBot allowed in robots.txt Not disallowed, explicitly or by wildcard

- OAI-SearchBot and ChatGPT-User allowed These are separate agents from GPTBot

- ClaudeBot and Claude-User allowed Also separate agents

- PerplexityBot allowed

- Google-Extended allowed Controls Gemini and AI Overviews use

- Page returns HTTP 200 to a bot user-agent Not 403 from a WAF or bot rule

- Content present in server HTML Over ~120 words of text without running JavaScript

- No login or consent wall on the content Interstitials are what the crawler caches

## Entity resolution — who and what is this?

Schema requirements depend on page type. Absence of inapplicable schema is not a defect.

- Valid JSON-LD present and parseable One syntax error voids the whole block

- Correct primary entity for the page type NewsArticle, Product, HowTo, JobPosting, etc.

- Organization schema present Anchors the publisher entity

- schema headline matches the visible H1 Word for word

- BreadcrumbList present Gives the engine site hierarchy

- Canonical URL declared

- about / mentions / sameAs used Disambiguates the entities you discuss

- Consistent entity naming sitewide Same name, description and category everywhere

## Extractability — can a model lift an answer?

This is where most otherwise-healthy pages lose citations.

- Key-facts list or table above the fold The densest extraction target on the page

- Core fact stated in the first 150 words Before the narrative wind-up

- Exactly one H1

- Three or more H2 subheadings Gives the model answer boundaries

- Question-shaped headings "What caused X?" beats "Background"

- Two or more lists Lifted cleanly; long prose is mangled

- Comparison table where relevant Reproduced near-verbatim

- Answer-first sections Heading, then the answer, then the detail

- Substantive length Under ~300 words is rarely cited

## Freshness and attribution — will they credit you?

Over 70% of AI-cited pages were updated within the last twelve months.

- datePublished present and machine-readable

- dateModified present and honest Undated content is treated as stale

- Visible date on the page Not just in the schema

- Named author, not "admin" or "team"

- Author bio with real credentials Linked profile or published work

- Person schema for the author

- Three or more external sources linked Cite the authorities you rely on

- Verbatim quotes in blockquotes Quoted statements get cited; paraphrase gets absorbed

- llms.txt published Low cost, still worth having

## Which of these checks matter most?

Three account for the majority of real-world failures. AI crawlers disallowed in robots.txt, usually by accident. Content that only exists after JavaScript runs, which is invisible in a browser and therefore rarely noticed. And a missing or stale dateModified, which quietly suppresses citation on any query where recency matters.

> The pattern in most audits is not that the content is weak. It is that something structural — a robots rule, a client-side render, an absent date — is preventing perfectly good content from being read at all.

## Does the checklist change by page type?

The blocking, extractability and freshness groups are universal. The entity-resolution group is not: the correct schema depends entirely on what the page is. Judging a news report against a product checklist produces a low score and advice that would make the page worse if followed.

Run this checklist automatically, scored against the profile for your page type.

Run a free AEO audit

## Related reading

- How to do an AEO audit — the manual process behind this list

- AEO audit vs SEO audit

- What is Answer Engine Optimization?