AEO Audit Checklist: 25 pass-or-fail checks for AI search readiness

An AEO checklist that only tells you what to do is not an audit. This is a 25-check AEO audit checklist built from pass or fail tests: run each one against a page you have already published, in under two minutes, and you get an honest answer on whether AI engines can access, answer with and trust it enough to cite.

I run SEO and GEO for clients from Pune, and I maintain this site's own AI-crawler access, schema and llms.txt file by hand. This guide groups 25 checks into seven areas: access, answerability, evidence, entity consistency, structured data, third-party presence and measurement. If the term itself is new to you, start with what AEO actually means before running the audit below.

What makes this an audit instead of a checklist

This AEO checklist works like an audit: every item is a pass or fail test against a page you have already published, not a task waiting to be done. My existing SEO and GEO checklist is the to-do version, the one you work through before or during publishing. This audit is what you run after, to check whether that work actually held up.

The difference matters because a checklist can be completed and still fail in practice. You can tick off "add FAQ schema" and still end up with an FAQ block that answers a question nobody searches for. An audit does not ask whether you did the task. It asks whether the result passes when actually tested.

One thing this audit does not do is promise a rank. There is no position one inside an AI answer the way there is in Google's organic results, so nothing here measures rank. What it measures is whether a page is eligible to be cited, linked or mentioned at all, which is the only goal an honest AEO audit can set.

How to score each check in under two minutes

Score every check pass, fail or not applicable, using the two-minute test listed next to it. Do not guess: open the actual page, the actual robots.txt file, or the actual Search Console report before marking anything. A guessed pass is worse than an honest fail, because it hides the real problem.

Add up your passes out of 25 and read the result against the table below. Treat "not applicable" checks, for example local business schema on a site with no physical location, as excluded from both the score and the total, not as automatic passes. If you want a fast on-page sanity check before you start, run the page through my free website SEO checker first.

Fix checks in the order they appear in this guide. A page that fails access never gets evaluated on answerability, evidence or anything else, so there is no point polishing an answer that AI crawlers cannot even reach.

Access checks: can AI systems reach your pages

Step 1: Run the access checks

Access checks confirm that AI crawlers and Google's own systems can actually fetch and read a page before you worry about anything else on it. Four checks live in this group, and each takes under two minutes with the right file already open in a browser tab.

Check 1: robots.txt allows the AI crawlers you actually want. Open yoursite.in/robots.txt and search for GPTBot, ChatGPT-User, ClaudeBot, PerplexityBot and Google-Extended by name. A pass means each one has an explicit Allow rule or no matching Disallow; a fail means a blanket rule sits above them that blocks every AI crawler along with regular search engines.

User-agent: GPTBot
Allow: /

User-agent: PerplexityBot
Allow: /

User-agent: ClaudeBot
Allow: /

Sitemap: https://yoursite.in/sitemap.xml

Check 2: you understand what each crawler actually does before you allow or block it. GPTBot collects training data for OpenAI's models, OAI-SearchBot surfaces pages inside ChatGPT's search feature, and ChatGPT-User fetches a page live when a user asks a direct question, according to OpenAI's crawler documentation. Perplexity splits the same way: PerplexityBot indexes for search results, while Perplexity-User fetches on a live user request and often ignores robots.txt entirely, per Perplexity's bot documentation. Blocking the wrong one removes you from citations; blocking the right one just opts out of training.

Check 3: you have not confused Google-Extended with Googlebot. Google-Extended controls whether Gemini apps and Vertex AI can use your content for training and grounding, and blocking it does not remove you from regular Search or from AI Overviews and AI Mode, which run on the same index Googlebot already builds. Google's own page on AI features and your website confirms that no separate opt-in or special markup exists for these features beyond normal Search access, at the time of writing (September 2026).

Check 4: the page renders without JavaScript. Fetch the live URL with a plain text tool, or use "view source" rather than the rendered DOM, and check whether your main content and headings appear in the raw HTML. A pass means the answer is readable without executing scripts; a fail means an AI crawler that does not render JavaScript sees a mostly blank shell, which is common on client-rendered sites without server-side rendering.

Answerability checks: does the page answer before you scroll

Step 2: Run the answerability checks

Answerability checks test whether a page gives an AI engine something quotable in the first few sentences, rather than making it dig for the point. This is the layer most people mean when they say GEO or AEO, and it is worth getting right.

Check 5: the opening paragraph answers the core question directly. Read only the first two or three sentences of the page and ask whether they alone would satisfy the search query. A pass reads like a direct briefing; a fail buries the answer under a story, a disclaimer or a company introduction first.

Check 6: headings are phrased as the questions people actually type. Scan every H2 and H3 on the page and check whether at least half could stand as a search query on their own, such as "how much does X cost" rather than "pricing". Question-shaped headings give an AI system a clean hook to match against a user's actual prompt.

Check 7: FAQ questions come from real searches, not guesses. Compare your FAQ block against Search Console query data, the People Also Ask box for your topic, or questions customers actually send on WhatsApp or email. A fail is an FAQ section invented to satisfy a schema requirement, answering questions nobody asked. My guide on whether FAQ schema helps you get cited in AI Overviews covers this in more depth.

Check 8: each H2 section could be lifted out and still make sense. Copy the text under one heading into a blank document with no other context and read it back. A pass reads as a complete, self-contained answer; a fail depends on a term or a reference explained three paragraphs earlier.

Evidence checks: author, dates and sources anyone can verify

Step 3: Run the evidence checks

Evidence checks test whether a page gives an AI system, and a skeptical reader, a reason to trust what it says. This is where E-E-A-T becomes something you can actually test, rather than a term you nod along to.

Check 9: a real author byline with a name and credential appears on the page, not just "Admin" or a company name. Scroll to the author section and confirm a full name, a role and at least one verifiable credential or link are visible, not only present in hidden schema.

Check 10: published and updated dates are visible and match the schema. Look for both dates in the page header, then view the page source and confirm datePublished and dateModified in the Article schema match what a reader sees. A schema date from last year on a page that reads as brand new is a common and avoidable fail.

Check 11: at least one factual claim links to a primary source. Scan the body for any number, policy or statistic and check whether it links to the organisation that actually published it, such as Google Search Central or a platform's own help centre, rather than to another blog repeating the same figure secondhand.

Check 12: at least one first-hand detail appears that could not be lifted from a competitor's page: a specific example, a number you measured yourself, or an observation from real work, rather than a summary of advice available on twenty other sites. Generic restatements are the easiest kind of page for an AI system to skip in favour of the original source.

Entity and structured data checks: does your identity match everywhere

Step 4: Run the entity and structured data checks

This step combines two related groups: entity consistency, meaning whether your identity matches everywhere it appears, and structured data validity, meaning whether the markup describing that identity is actually correct. My schema markup guide covers the basics if any of this is unfamiliar. Six checks live here.

Check 13: your business name, address and phone number are identical on your website, Google Business Profile and every directory listing you control. Pull up all three side by side and compare them character by character, since "Road" versus "Rd" or a missing suite number is enough to weaken the match. My guide on NAP consistency in local SEO covers this check in more depth if you find a mismatch.

Check 14: your Person or Organization schema matches the visible author bio exactly. Open the page source, find the author object in the Article schema, and compare the name, job title and credentials against what the visible bio actually says. A fail is schema that lists a credential or title nowhere shown to a human reader, which looks inflated the moment a system cross-checks it.

Check 15: sameAs links in your schema point to real, active profiles you control. Click every URL listed under sameAs, whether that is LinkedIn, Instagram or another platform, and confirm each one loads and is actually yours. A dead link or an unclaimed profile undermines the entity match the schema is trying to build.

Check 16: Article or BlogPosting schema validates with zero errors. Paste the live URL into Google's Rich Results Test and read the result. A pass shows the detected schema types with no errors listed; warnings are usually fine to leave, but any error means a field is malformed and the markup may be ignored entirely.

Check 17: FAQPage schema, if you use it, matches the visible question and answer text word for word. Compare the schema JSON against the rendered page. A mismatch between what a user reads and what the schema claims erodes trust in every other schema block on the same page, not just the FAQ.

Check 18: LocalBusiness or Organization schema appears sitewide, not only on the homepage. Spot-check three or four interior pages, such as a blog post or a service page, and confirm the same business schema is present. My LocalBusiness schema guide has a full copy-paste example if this check fails.

Third-party presence checks: what the rest of the web says about you

Step 5: Run the third-party presence checks

AI systems weigh more than your own site when they decide whether to trust it, so this step looks outward instead of at your own pages. Three checks live here, and say a dental clinic in Baner should expect the first two to take longer than the on-site checks above.

Check 19: your business appears in a handful of listings you do not control directly. Search your business name in quotes and check whether directories, industry associations or local publications mention you beyond your own website and social profiles. A single source describing your business is easy for any system to discount; several independent ones are harder to ignore.

Check 20: recent, visible reviews exist somewhere a person or a crawler can read them. Check your Google Business Profile or the review platform relevant to your industry for reviews within the last few months, not just a cluster from years ago. A pass is recent activity; a fail is a profile that looks abandoned even if the historical rating is good.

Check 21: at least one mention or link comes from a source you genuinely do not own. Search for your business name or domain and look past your own social accounts for a guest article, a podcast appearance, a partner page or a press mention. This is the slowest check to pass honestly, and there is no shortcut that does not involve someone else choosing to mention you.

Measurement checks: proving any of this is working

Step 6: Set up the measurement checks

The last four checks are not about a single page. They test whether you have any way of knowing if the previous 21 checks actually change anything. For the tools that make these four checks easier to run every month, see my AEO tools guide, the companion piece to this audit.

Check 22: GA4 has a channel or segment that isolates AI referral traffic from generic referral traffic. Without this, a visit from a ChatGPT or Perplexity citation gets buried inside a general "Referral" bucket alongside every other external link, and you lose the ability to see it at all. My guide on tracking AI traffic in GA4 walks through building this channel step by step.

Check 23: you check both AI-relevant Search Console reports at least monthly. The dedicated Generative AI performance report shows impressions only, with no clicks or position, for links shown inside AI Overviews and AI Mode, and Google says it reached every website worldwide as of 31 August 2026, though your property may still show no data if your impressions there are too low. Separately, the general Performance report still folds AI-feature clicks into its existing Web totals rather than breaking them out, per Google's AI features and your website page. Read both in my Search Console for beginners guide if you have not set them up yet.

Check 24: a monthly prompt-test log exists somewhere, even a simple spreadsheet. Ask the same three to five real questions in ChatGPT, Perplexity or Google's AI Mode every month, and record whether your business or page shows up, gets linked, or is absent. This is manual and slightly tedious, and it is also the only way to see a trend rather than a single snapshot.

Check 25: you have a place to log a citation when you actually see one, without overstating it. A pass is a habit of noting the date, the platform and the exact prompt when your page gets mentioned; a fail is either never checking, or assuming a citation from indirect evidence like a traffic bump alone, since AI referral traffic is still hard to attribute with full confidence.

How this site scores on the checks I can prove

I will not claim results I cannot show you, so this section only covers checks I can point to directly rather than describe from memory.

This site's robots.txt passes Check 1: it explicitly allows GPTBot, ChatGPT-User, ClaudeBot, PerplexityBot, Google-Extended and over a dozen other named AI crawlers, with a Sitemap line at the end rather than a blanket block. It pings IndexNow after every deploy too, so new and updated pages reach Bing and other participating engines faster, a genuine access aid even though it is not one of the 25 checks above.

The site also keeps an optional llms.txt file at the domain root, purely as documentation for AI systems that choose to read it. I am not scoring that as a pass on any check here, because Google's own guide to optimizing for its AI features states plainly that Google Search ignores llms.txt files, so they neither help nor harm visibility there. I keep the file anyway on the chance another AI system finds it useful, not because it earns anything with Google.

On structured data, every guide on this site ships Article, FAQPage, Person, BreadcrumbList and Speakable schema by default, including this one, which covers most of Check 16 through Check 18 by default rather than by manual effort on each new page.

What I am not claiming is that any of this has produced a confirmed citation inside ChatGPT, Perplexity or Google AI Mode. I have not tracked that consistently enough to state it as fact, which is itself a fail on Check 24 that I am fixing by building the monthly log this audit recommends.

Common mistakes when running an AEO audit

The most common mistake is running the answerability checks before the access checks pass. Polishing an opening paragraph on a page that robots.txt is quietly blocking wastes the effort twice: once on the writing, and again when you eventually find the block.

A second mistake is treating a single pass as permanent. A template change, a migration or a new plugin can silently reintroduce a JavaScript-only render or drop a schema block, so a check that passed in March can fail by June without anyone touching the content itself.

A third mistake is inventing FAQ questions to satisfy Check 7 instead of pulling them from real searches. An AI system has no reason to prefer a page answering a question nobody asked over one answering a question people actually type, so invented FAQs pass the schema validator and fail the check that actually matters.

When an AEO audit is not worth running yet

If your site fails basic technical SEO, this audit is premature. A page stuck in "Crawled, currently not indexed" in Search Console has bigger problems than answerability, and no amount of AEO polish fixes an indexing failure. Run a free SEO audit first.

If you have fewer than a handful of pages that already rank anywhere on page one or two for their target query, spend that time on content and links instead. AEO checks assume a page with some existing visibility; they will not turn an invisible page into a cited one on their own.

And if your business genuinely has no online presence beyond the website itself, such as a brand-new practice with no reviews or listings yet, several third-party presence checks will fail no matter what you do to the site, simply because there has not been time to build them. That is a timing problem, not a technical one, and it resolves on its own as the business ages. For the bigger-picture plan to work toward once your pages pass, see my AEO strategy guide.

Frequently asked questions

What is an AEO checklist?

An AEO checklist is a set of checks that make content easier for AI answer engines to access, understand and cite. This audit version goes further: instead of listing tasks to complete, it gives each check a pass or fail test you can run in under two minutes against a page you already have live. The result tells you exactly which of the 25 checks your site currently passes, rather than handing you a general to-do list of best practices to work through blind.

Can I rank number one in ChatGPT or Google AI Mode?

No. There is no ranking position inside an AI answer the way there is in Google's ten blue links. AI engines either mention your business, cite your page with a link, or leave you out entirely, and the same prompt can return a different mix of sources on separate runs. The realistic goal of an AEO audit is to raise the odds of being cited, linked or mentioned, not to chase a rank that does not exist in this format.

Do I need special schema markup to appear in AI Overviews or AI Mode?

No. Google's own documentation on AI features states that no special markup or additional files are required to appear in AI Overviews or AI Mode, and that normal SEO fundamentals apply. Structured data such as FAQPage or Article schema still helps by making a page's structure unambiguous to any crawler, but it works as a clarity aid, not an entry requirement. Skipping schema will not by itself disqualify a well-written, accessible page from being cited.

How is an AEO audit different from an SEO audit checklist?

An SEO audit checks whether Google can crawl, index and rank a page at all. An AEO audit checks a narrower, AI-specific layer on top of that: whether AI crawlers are allowed in, whether the page opens with a direct answer, whether the author and sources are verifiable, and whether your business identity is consistent across the web. Most AEO checks assume the SEO foundation already passes, so run a technical SEO audit first if you have not already.

How often should I re-run this AEO audit?

Run the full 25-check audit quarterly, and repeat the access and structured data checks any time you change hosting, redesign a template or migrate a CMS, since those changes are the most common cause of a page silently failing. AI products and crawler policies also change often enough that a pass six months ago is not a guarantee today. Treat this as a recurring habit rather than a one-time project, the same way you would monitor Search Console.

Related guides

• AEO tools guide: free and paid ways to track AI visibility
• Free SEO and GEO checklist for 2026
• What is AEO (answer engine optimization)?
• Does an FAQ section help you get cited in AI Overviews?
• What is robots.txt in SEO?
• Schema markup guide for beginners
• LocalBusiness schema markup: JSON-LD examples
• Google Search Console for beginners

Your next step

Start with the access checks: they take ten minutes and catch the failures that make every other check pointless. Once a page passes access and answerability, read the AEO tools guide to set up free monitoring, or get in touch if you want a second pair of eyes on the results.