An AEO audit should leave you with findings you can verify. Check whether the intended public page can be accessed, whether its answer is clear and whether its claims have evidence. Keep search access, training permissions and content quality as separate decisions instead of treating one crawler setting as a complete AI visibility strategy.
TLDR: Review indexability, crawler access, any available Search Console AI inclusion control, rendered text, source support, internal links and measurement. Document the purpose of each control before proposing a change. Do not assume a special schema or AI text file guarantees inclusion in an answer.
Sources checked September 7, 2026. The policy examples are fictional review decisions, not instructions to deploy a robots.txt change. No crawl settings or live pages were changed for this article.
Start with the question and the intended page
Choose a buyer question and the page that should answer it. Record the intended audience, the claim you want the page to support and the evidence currently available. A site-wide audit can then group findings around specific page purposes.
We recommend reviewing the existing content before creating another article. If a suitable page already owns the question, improve its clarity or evidence rather than adding a near-duplicate answer. Keep the proposed action and its rationale in a review log.
Our KPI gap-analysis guide provides related measurement context. This audit focuses on access and answer evidence, not a new composite visibility score.
Check search eligibility and applicable AI controls
Google states that supporting pages in AI Overviews and AI Mode must be indexed and eligible for a search snippet. It does not require special AI files or a special schema type for those features, and meeting the requirements does not guarantee inclusion. Google: AI features and your website.
Google's newer guidance also points to a Search Console inclusion control for generative AI features. The linked help page describes a rollout to a subset of website owners, with include, exclude and inherited choices. Check whether the control is available for your property and record the applicable setting without changing it as part of the audit. Google: Search generative AI control.
That control's documentation distinguishes Search generative visibility from AI-training choices and general search exclusion. Do not infer that a crawler allow rule proves your property is included. Review each applicable control separately and obtain approval before any change. Google: control scope and limitations.
For each intended public page, we recommend checking its response, canonical, indexing directives and links from relevant pages. Inspect the content that a crawler can receive as well as the visual design. Record any differences between staged content and the public version.
Do not mark a page indexed merely because it loads or appears in a sitemap. Keep public delivery, search-engine indexing and observed answer inclusion as separate status fields.
Check AI crawler access separately from training access
OpenAI documents OAI-SearchBot for search and GPTBot for content that may be used to train foundation models. Their settings are independent. The documentation also distinguishes user-initiated ChatGPT-User access from automatic search crawling. Review the intended purpose before approving a policy change. OpenAI: crawler overview.
The following fictional policy review illustrates the questions to resolve. It is not a recommended default for every website.
| Business decision | What the reviewer should verify | Approval needed before change |
|---|---|---|
| Allow a public article to be considered for search | Relevant search crawler policy and actual delivery | Website and content owner |
| Restrict use for model training | The provider's documented training control | Authorized policy owner |
| Protect a private customer workspace | Authentication and authorization boundaries | Security and application owner |
| Diagnose a blocked request | Exact response, infrastructure rule and verified request context | Owner of the affected control |
We recommend keeping private content behind appropriate access controls. Do not treat an exclusion instruction to a crawler as the security boundary for confidential information. Any change to a firewall, CDN rule or robots policy should have a scoped reason, approval and rollback record.
Review the answer as a reader would
Find the sentence that directly answers the target question. Check whether it states the relevant condition, limitation and next step. Then inspect the support for every factual claim that follows.
We recommend an editorial checklist covering:
- Direct answer: A reader can identify the answer without decoding an introduction.
- Evidence: Product facts link to current primary sources; internal claims have approved proof.
- Scope: Plan, date, audience and other limitations are visible where they matter.
- Examples: Synthetic cases are labeled and do not imply observed client results.
- Navigation: Relevant internal links help the reader continue without duplicating the same destination repeatedly.
- Consistency: Metadata, visible content and any structured data say the same thing.
These are editorial review criteria, not a guaranteed method for earning an AI citation. If a claim cannot be supported, narrow it, remove it or hold it for review.
Inspect rendered answers, schema markup and destinations
Verify that important text is present in the delivered page and that the links lead to the intended content. Check mobile readability, table behavior and image alternatives. A screenshot of an answer inside an image should not be the only usable version of the explanation.
We recommend checking the destination of each citation and CTA as part of the same review. A resource mentioned as available needs a working public delivery path. A planned download should remain in the internal backlog until released.
If your content program uses a broader loop of research and revision, our Loop Marketing article provides context. Preserve the evidence behind each change so the next review can distinguish an assumption from an observation.
Produce an actionable findings log
Record the URL, question, observed problem, evidence, proposed action, owner and verification condition. Separate a confirmed failure from an untested possibility. For example, “the public destination returns an error” is a different finding from “this page may need stronger evidence.”
Prioritize failures that prevent access or mislead the reader before speculative optimization. After an approved repair, repeat the relevant check and append the result. Do not rewrite the original finding as though the issue never existed.
Organize the AEO audit around verifiable layers
Answer engine optimization adds questions about source use and answer visibility to an ordinary technical and editorial review. It should not replace SEO or turn every missing citation into a crawl problem.
Use a consistent audit methodology across priority pages. Record the buyer question, page purpose, technical observations, content evidence and measurement scope. Keep each finding tied to the layer it actually concerns.
Technical access and indexability
Check the page response, canonical URL, indexing directives and applicable crawler policy. Inspect any relevant infrastructure control using verified request information.
A blocked search crawler and a private application are different situations. The intended public article may need a technical correction; the private application needs to remain protected.
Keep changes to crawler access separate from the read-only audit. Record the authorized correction and retest the actual destination after it is applied.
Content quality and entity clarity
A reader should be able to identify the organization, product or service being discussed. Use accurate names and explain terms that could be confused with another feature or company.
Check whether the page answers its stated question with supported detail. A broad introduction may need a clearer answer; an answer without material qualifications may need a more careful one.
Do not create a new claim about credentials, clients or results solely to increase perceived authority. Retain evidence for the claims the organization can make.
Internal linking and duplicate content
Find the page that owns the buyer decision and inspect the relevant internal links. A related article should provide a useful next step rather than repeat the same explanation under another URL.
Where several pages overlap, record the intended role of each before proposing a merge or redirect. Content optimization should preserve useful material and avoid unapproved URL changes.
A page being easy to reach from the site is an observable condition. That does not prove an AI system will select it for a response.
Structured data validation
Review the structured data actually delivered with the page. Its type, headline, author and other included facts should agree with the visible content and documented implementation.
Use the appropriate validation tool for the supported markup. Passing validation is evidence about the markup checked, not a promise of rich results or an AI citation.
Do not add Organization schema, FAQ markup or a special AI file merely because an audit tool suggests it without a relevant requirement.
Keep AI response observations separate from page defects
Citation tracking can help identify answers worth inspecting. Preserve the prompt, platform, observed response, linked sources and date.
An AI-generated answer may omit your brand even when the page is accessible and useful. That absence is an observation within the selected scope, not proof that the website has a technical defect.
Review the cited evidence before proposing a change
Open the actual sources and compare them with the answer's claim. Note current product limits, publication evidence and any unsupported conclusion.
If the source answers a different question, do not use it as a model for your article. If it supports a useful missing detail, verify that detail through the appropriate primary source.
Keep competitors' observations and your proposed changes distinct. A competitor appearing in an answer does not establish what caused the selection.
Do not turn every AI platform into the same test
Different AI assistants and search surfaces expose different settings and evidence. Record the system and visible mode rather than generalize from a single tool.
Google AI Overviews, a conversational search answer and a user-initiated browser request should not be treated as identical crawler events. Use the provider's current documentation for the control being reviewed.
Prioritize the audit findings
Fix confirmed access failures and misleading claims before speculative formatting changes. A broken CTA or an obsolete entitlement statement is a concrete issue even without a visibility forecast.
Give each finding a proposed action, owner and completion condition. Keep confirmed defects, research hypotheses and unavailable evidence in separate states.
After a correction, repeat the relevant technical or editorial check. Measure search visibility and subsequent AI responses separately, using the same scope where possible.
Keep the AEO audit connected to technical SEO
Start with access and delivery. If a page cannot be fetched or understood by the intended search engine, adding answer sections will not fix that technical defect. Review AI crawler access separately from the accuracy and usefulness of the published answer.
A traditional SEO audit and an AEO audit share questions about page availability, canonical URLs, internal links and supported content. The AEO review adds the observed answer, its cited sources and whether those sources support the statement. It does not replace the ordinary SEO checks.
Keep schema markup consistent with visible content. Do not add a special AI claim or unsupported rich-result promise to make the audit look complete. Record an issue, its evidence, its owner and the test that will verify the correction.
Frequently asked questions
Do we need a special AI schema to pass this audit?
Do not add markup solely because it is described as an AI requirement. Verify the search provider's current guidance and make any structured data match the visible content and applicable type.
Does allowing a search crawler guarantee a citation?
No. Access is one condition to investigate. It does not establish indexing, selection or inclusion in a particular answer.
Can a robots.txt rule protect private customer information?
We recommend using appropriate authentication and authorization for private content. Treat crawler preferences as a separate policy decision, not a substitute for an access boundary.
What should the audit deliver?
A traceable list of observed findings, evidence, responsible owners and specific verification conditions. Keep untested hypotheses and proposed live changes clearly labeled.
Make AEO work auditable
Our consulting services can help you review the page, evidence and technical conditions behind a target question.
Talk with Selworthy with the pages and questions you want to evaluate before changing crawler policies or publishing new claims.
