Skip to main content Scroll Top

AEO Audit: The Full Checklist, Published

Updated September 2026 · Written and maintained by the Progression Agency strategy team

An AEO audit is eight checks that establish whether AI answer engines can reach, read and quote your site — including the three you can run yourself this afternoon, with the commands.

On this page · 13 sections
  1. What an AEO audit is
  2. Run the first three checks yourself, today
  3. The full checklist, in the order that matters
  4. What audits actually find
  5. What to fix first
  6. What a weak AEO audit looks like
  7. What you should receive
  8. Do you need a full audit, or just the free checks?
  9. Everything else we have written on search, AI and getting found
  10. The eight checks as a table you can work through
  11. The five crawlers to check by name
  12. Audit outputs, and how a second opinion could verify each one
  13. How this differs from an SEO audit

The short answerAn AEO audit answers one question: can AI answer engines reach your pages, read them without running JavaScript, find a clean answer inside them, and trust what they find. Eight checks, about a working week, published at $1,000 to $4,000. The first three cost nothing and find most of what is wrong on most sites: read robots.txt for the five AI crawlers, fetch a page with JavaScript disabled, and confirm AI user agents get a 200 rather than a 403. The whole method is on this page.

The AI search visibility audit in 15 seconds: how AI assistants answer without you, what the audit measures and what changes afterward.

Failure proportions cited here are our own observation across our own audits, labelled as such rather than presented as industry research. Crawler names current September 2026.

The AEO audit, in the order you should run itThe AEO audit, in the order you should run it
Steps one to three take an afternoon and find most of what is wrong on most sites. Everything after that is refinement.

What an AEO audit is

A short, mostly technical review that answers one question: can AI answer engines reach your pages, read them, find a clean answer inside them, and trust what they find. Eight checks, roughly a working week, and the published price band for the work is $1,000 to $4,000.

It is not a content quality review. It is not a score out of a hundred. Most of it is reading configuration files and comparing two versions of the same page, and the first three checks find most of what is wrong on most sites.

We publish the whole method below, including the commands, because the method is not the valuable part. Knowing which finding is worth its cost to fix is.

robots.txt — Check. Five crawlers by name. Free. Two minutes.
Scriptless render — Check. The expensive finding, found cheaply.
Status codes — Check. 403s to AI bots are invisible in a browser.
Log grep — Check. Turns theory into evidence.
Passage structure — Check. Does the section answer its own heading.
Schema vs copy — Check. Contradiction is worse than absence.

What it checks

Crawler permissions, status codes returned to AI user agents, whether pages render without JavaScript, whether AI crawlers actually arrive, passage structure, schema agreement, entity consistency, and whether any baseline exists.

What it produces

Six outputs, each a file you keep: a crawler verdict per bot, a render table, log evidence, structure findings, a schema diff, and a dated prompt baseline.

What it costs

Published at $1,000 to $4,000 for 10 to 20 hours. The first three checks cost nothing and you can run them yourself this afternoon.

What it is not

A rewrite plan, a keyword map, or a recommendation to buy a retainer. A competent audit sometimes concludes that nothing further is needed this year.

Run the first three checks yourself, today

Open your robots.txt and search for the five AI crawlers by name. Fetch a commercial page with JavaScript disabled and see how much text remains. Request that page with an AI user agent string and confirm you get a 200 rather than a 403.

That is about an hour of work, it requires no tools and no budget, and on a meaningful minority of sites it finds a complete explanation for why the brand is absent from AI answers.

How to run the first three checks yourself, todayHow to run the first three checks yourself, today
We publish this because the method is not the valuable part. Interpreting what you find, and knowing which fix is worth its cost, is.
GPTBot — Crawler. OpenAI crawling for training and indexing.
OAI-SearchBot — Crawler. OpenAI crawling for search results.
ClaudeBot — Crawler. Anthropic's crawler.
PerplexityBot — Crawler. Perplexity crawling and live fetch.
Google-Extended — Crawler. Gemini grounding and AI training use.
Bingbot — Crawler. Still underpins several AI surfaces.

Check one: robots.txt

Visit yoursite.com/robots.txt and search for GPTBot, OAI-SearchBot, ClaudeBot, PerplexityBot and Google-Extended. A Disallow under any of them is a complete answer for that surface, and it is very often there by accident.

Check two: rendering

Disable JavaScript in your browser’s developer tools and reload a commercial page, or fetch it with curl. If what remains is a navigation bar and a footer, rendering is your dominant problem and everything else is secondary.

Check three: status codes

Request the page with an AI crawler user agent string. A 403 or a soft 404 returned only to bots is invisible in normal browsing and completely disqualifying.

What to do with what you find

Write it down with the date. That is now a rudimentary baseline, it cost nothing, and it is more than ninety per cent of sites have.

The full checklist, in the order that matters

Eight checks, ordered so that the cheapest and most decisive come first. If check one fails, nothing further matters until it is fixed, because a blocked crawler makes every other consideration hypothetical.

The ordering is the part most audits get wrong. Beginning with content quality on a site that returns nothing to a scriptless fetch is analysing a page nobody is reading.

The pass/fail checklistThe pass/fail checklist
The two ‘part’ rows are conditional: llms.txt is hygiene rather than leverage, and accordions are fine if the content is in the HTML rather than injected on click.
How long each part takesHow long each part takes
One working week for a mid-sized site. The published price band for this work is $1,000 to $4,000, which is consistent with 10 to 20 hours.

What audits actually find

Across the audits we have run, weak passage structure and a missing baseline are near-universal, entity inconsistency and schema contradictions are common, JavaScript dependence affects a substantial minority, and an outright blocked crawler turns up in roughly a third of cases.

Those proportions are our own observation on our own sample, stated as such. They are not an industry statistic and we would rather label them honestly than dress them up as research.

What we actually find, across auditsWhat we actually find, across audits
These are the proportions we see, stated as our own observation rather than as an industry statistic. The 92% is the one that matters most: almost nobody has a baseline, so almost nobody can prove anything.
Blocked crawler — Finding. Free to fix, total absence if unfixed.
JS-only content — Finding. Expensive, and the commonest severe failure.
Answer in paragraph four — Finding. Cheap to fix, large effect on extraction.
Schema contradicts page — Finding. Cheap to fix, erodes trust when wrong.
Sources disagree — Finding. Slow to fix, causes wrong answers about you.
No baseline — Finding. Free to fix, and nothing is provable without it.

Almost nobody has a baseline

This is the finding that surprises clients least and matters most. Without a dated prompt set recorded verbatim, every subsequent claim about improvement is unfalsifiable.

Blocked crawlers are usually accidental

A security plugin default, an inherited robots.txt, a WAF rule added for a different reason. Almost nobody blocks GPTBot deliberately and then wonders why they are absent.

JavaScript dependence is invisible until tested

The page looks perfect to everyone who checks it, because everyone checks it in a browser. Only a scriptless fetch reveals it.

Passage structure is the quiet one

Not dramatic, not free, but it affects every page and the fix improves the page for human readers too.

What to fix first

Order by consequence divided by effort. A blocked crawler is minutes to fix and total absence if you do not, so it always comes first. JavaScript dependence is severe and expensive, so it is the decision that shapes the budget.

Everything else is a judgement call, and a good audit makes that judgement explicitly rather than handing you a list of twenty findings in no particular order.

What to fix first, by effort and consequenceWhat to fix first, by effort and consequence
Top-left is where to start every time: a blocked crawler costs minutes to fix and total absence if you do not.
Effort versus impact, by audit stepEffort versus impact, by audit step
The first two bars are the argument for doing this at all: near-zero effort, near-maximum impact. Entity work is the inverse — high effort, moderate payoff, but it is what stops assistants being wrong about you.
Fix access — Next. Free, minutes, binary.
Decide on rendering — Next. Engineering call, biggest variable.
Restructure key templates — Next. Cheap, high extraction gain.
Reconcile schema — Next. Cheap, protects accuracy.
Unify the description — Next. Slow, stops wrong answers.
Re-run the prompt set — Next. Monthly, identical wording.

What a weak AEO audit looks like

Five warning signs, and any two together mean you have bought a content review with AI in the title: a score out of a hundred with no stated method, llms.txt as the headline finding, rank data presented as evidence of AI visibility, no request for server logs, and a recommendation to buy a retainer that clearly preceded the work.

The last is the most common. An audit whose conclusion was decided before the first check was run is a sales document, and it is usually recognisable because none of its findings are specific to your site.

What a weak AEO audit looks likeWhat a weak AEO audit looks like
Five reds. Any two of them together mean the audit was a content review with AI in the title.
Score out of 100 — Red flag. No stated method behind the number.
llms.txt headline — Red flag. A checklist, not an audit.
Rank data as proof — Red flag. Cannot see inside an answer.
No log request — Red flag. Crawler claims become unprovable.
Retainer regardless — Red flag. The conclusion preceded the work.
Content review only — Red flag. The wrong discipline entirely.

What you should receive

Six outputs, each independently checkable: a crawler verdict per bot with the exact lines to change, a render table of scripted versus scriptless word counts per template, log evidence of which bots arrived, structure findings naming specific sections, a schema diff, and a prompt baseline with verbatim answers and a date.

If any of those arrives as a narrative rather than as data, ask for the data. All six are reproducible, which means a second supplier could verify them, which is the point.

Crawler verdict — Output. Allowed or blocked, per bot, with the exact lines.
Render table — Output. Scripted vs scriptless word counts per template.
Log evidence — Output. Which bots arrived, when, with what status.
Structure findings — Output. Sections whose answer is not in sentence one.
Schema diff — Output. Claims present in markup but absent from copy.
Prompt baseline — Output. Verbatim answers, dated, reproducible.
The numbers an AEO audit turns upThe numbers an AEO audit turns up
The zero and the two together are the case for auditing before buying anything else.

Do you need a full audit, or just the free checks?

If your site is simple, your buyers rarely use assistants, and the free checks come back clean, record a baseline and stop. That is a real recommendation and we make it regularly.

A full audit earns its fee when the site is complex or app-like, when your category is one where buyers genuinely ask assistants for recommendations, or when the free checks turn up something you cannot interpret on your own.

Who needs a full audit versus a quick checkWho needs a full audit versus a quick check
Bottom-left is a real recommendation: run the free checks, record a baseline, and do nothing else this year.

Quick check only

Simple site, brand-led demand. An hour of your time, no budget.

Technical audit only

Complex or app-like site, but assistants are not yet how your buyers research. Fix access and rendering, skip the rest.

Baseline only

Clean site, low urgency. Record the prompt set, re-run it monthly, act when competitors start appearing and you do not.

Full audit

Complex site and question-led demand. All eight checks, ending in a scoped programme with prices attached to specific findings.

Audit plus engineering review

App-like site with severe rendering problems. The rendering conversation needs your developers in the room, and it should happen before anything is signed.

Want the interpretation rather than just the checklist?

The method above is free and complete. What an audit adds is knowing which finding is worth its cost on your site, and a baseline recorded properly so everything afterwards can be proven.

/ai-visibility-audit

Want this done for your site?We run all eight checks in about a working week and publish the price before we start.

Get a free proposal

Everything else we have written on search, AI and getting found

AI, AEO and what is changing

Websites and design

Choosing and working with an agency

Social, content and brand

By industry and by situation

Not sure which of these applies to you?Tell us the situation and we will say plainly what we would do first, and what we would not.

Talk it through

The eight checks as a table you can work through

The full audit checklist, in the order we run it, with what each check is looking for.

The AEO audit checklist
#CheckHow to run itFail conditionCost to fix
1Crawler permissionsRead /robots.txt for the five AI agentsAny Disallow under an AI agentFree
2Scriptless renderingFetch the page with JavaScript disabledLittle or no body text returnedHigh
3Status codesRequest with an AI user agent string403, 401 or a soft 404Low
4Crawler arrivalGrep server logs for AI user agentsNo hits in the last 30 daysDepends on 1-3
5Passage structureRead each section against its headingAnswer is not in the first sentenceMedium
6Schema agreementValidate, then compare with visible copyMarkup claims what the page omitsLow
7Entity consistencyList every source describing the businessDescriptions disagreeMedium
8BaselineRun a fixed prompt set, store answers verbatimNo dated baseline existsFree

The five crawlers to check by name

Five user agents to test explicitly, because a catch-all rule tells you nothing about how each is treated.

Who to look for in robots.txt and in your logs
User agentOperated byWhat it governsConsequence of blocking
GPTBotOpenAICrawling for training and indexingReduced presence across OpenAI surfaces
OAI-SearchBotOpenAICrawling for search resultsAbsent from ChatGPT search
ClaudeBotAnthropicCrawling for ClaudeAbsent when Claude searches the web
PerplexityBotPerplexityCrawling and live fetchAbsent from Perplexity answers
Google-ExtendedGoogleGemini grounding, AI training useExcluded from Gemini grounding

Audit outputs, and how a second opinion could verify each one

Every output we hand over, with the method someone else could use to check it independently.

Every output is reproducible on purpose
OutputWhat it containsHow anyone else could check it
Crawler verdictAllowed or blocked per bot, with exact linesOpen robots.txt and read it
Render tableScripted vs scriptless word counts per templatecurl the URL and count the words
Log evidenceWhich bots arrived, when, with what statusRe-run the same filter on the logs
Structure findingsNamed sections whose answer is not in sentence oneRead the section
Schema diffClaims in markup absent from visible copyRun any schema validator
Prompt baselineFixed prompts, verbatim answers, datedAsk the same prompts yourself

How this differs from an SEO audit

The overlap and the genuine differences, stated so you are not paying twice for the same work.

Same word, different discipline
DimensionSEO auditAEO audit
Primary questionCan this rank among linksCan this be reached, read and quoted
Crawlers consideredGooglebot, BingbotPlus GPTBot, OAI-SearchBot, ClaudeBot, PerplexityBot, Google-Extended
JavaScriptUsually assumed renderedAssumed not executed
Content focusKeyword coverage and depthWhether a section answers its own heading
Structured dataPresent and validPresent, valid and agreeing with the copy
EvidenceCrawl tool exportServer logs and scriptless fetches
Deliverable ends withA prioritised fix listA fix list plus a dated prompt baseline
Typical duration2-4 weeks1 week

Getting found in search

AI, AEO and what is changing

Paid media and lead generation

Websites and design

Choosing and working with an agency

Social, content and brand

By industry and by situation

Frequently asked questions

What is an AEO audit?
A short technical review that answers whether AI answer engines can reach your pages, read them, find a clean answer inside them and trust it. Eight checks, roughly a working week, published at $1,000-$4,000 for 10-20 hours.
What does an AEO audit check?
Crawler permissions, status codes returned to AI user agents, rendering without JavaScript, whether AI crawlers actually arrive, passage structure, schema agreement with visible copy, entity consistency, and whether any baseline exists.
Can I run an AEO audit myself?
The first three checks, yes, in about an hour with no tools. Read robots.txt for the five AI crawlers, fetch a commercial page with JavaScript disabled, and request it with an AI user agent to confirm a 200. The full method is published on this page.
How much does an AEO audit cost?
Published at $1,000-$4,000 for 10-20 hours of work. The three highest-impact checks cost nothing and you can run them today.
How long does an AEO audit take?
About one working week for a mid-sized site. Hour one settles the crawler question; hours two to four settle rendering; the rest is structure, schema, entity and baseline.
What is the most common AEO audit finding?
No prompt baseline exists. Almost nobody has one, which means almost nobody can prove whether anything they did afterwards worked.
What is the most severe finding?
Content that only exists after JavaScript runs. It looks perfect in a browser and returns almost nothing to the crawlers that matter, and fixing it is an engineering project.
What is the cheapest finding to fix?
A blocked AI crawler. It costs nothing, takes minutes, and it is the difference between being a candidate for citation and being absent entirely.
Which crawlers should an AEO audit check?
GPTBot and OAI-SearchBot from OpenAI, ClaudeBot from Anthropic, PerplexityBot, and Google-Extended, which governs Gemini grounding and AI training use.
How do I check if GPTBot can reach my site?
Read robots.txt for a Disallow under GPTBot, request a page with the GPTBot user agent string and confirm a 200 response, then grep your server logs for GPTBot hits in the last 30 days.
Is an AEO audit the same as an SEO audit?
No. An SEO audit asks whether a page can rank among links. An AEO audit asks whether it can be reached, parsed without JavaScript, and quoted. The evidence is server logs and scriptless fetches rather than a crawl tool export.
What should an AEO audit deliver?
Six reproducible outputs: a crawler verdict per bot, a render table of scripted versus scriptless word counts, log evidence, structure findings naming specific sections, a schema diff, and a dated prompt baseline with verbatim answers.
What is a prompt baseline?
A fixed set of buyer questions run once, with the answers stored verbatim and dated. Because answers vary between runs, it is the only way to distinguish a real change from ordinary variance later.
How do I know an AEO audit is any good?
It reads your server logs, fetches pages with scripts disabled, names the crawlers it tested, stores verbatim answers, and ends with findings specific to your site rather than a generic recommendation to buy a retainer.
What are the warning signs of a weak AEO audit?
A score out of 100 with no stated method, llms.txt as the headline finding, rank data offered as proof of AI visibility, no request for logs, and a retainer recommendation that clearly preceded the work.
Should the audit include llms.txt?
It can note it, but it should not lead with it. Adoption is low and it is not a ranking factor. Treat it as hygiene.
Do I need an audit if my site is simple?
Probably not. Run the three free checks, record a baseline, and act only if they turn up something or if competitors start appearing in answers where you do not.
What happens after the audit?
Fix access first because it is free, decide on rendering because it is the biggest variable, then restructure key templates, reconcile schema, unify the business description, and re-run the prompt set monthly.
Can an AEO audit conclude I need nothing?
Yes, and it does regularly — most often for clean server-rendered sites in categories where buyers still search rather than ask.
Why do you publish the whole method?
Because the method is not the valuable part. Knowing which finding is worth its cost on your particular site, and recording a baseline properly, is what an audit actually adds.
What is an AEO report?
The written output of the audit: the six deliverables above, with findings prioritised by consequence divided by effort rather than listed in no particular order.
How often should the audit be repeated?
The full audit annually, or after any significant site change. The prompt baseline should be re-run monthly, because that is the part that tracks movement.

Want this done for your site?We build and maintain the search, content and paid programmes described on this page.

Get a free proposal

Get a free marketing proposal

Tell us what you are trying to grow and we will come back with a plan, not a pitch deck. Same-day reply on weekdays.

Our website uses cookies from third party services to improve your browsing experience. Read more about this and how you can control cookies by clicking "Privacy Preferences".
Privacy Preferences
When you visit our website, it may store information through your browser from specific services, usually in form of cookies. Here you can change your privacy preferences. Please note that blocking some types of cookies may impact your experience on our website and the services we offer.
Privacy Policy
You have read and agreed to our privacy policy
Required
Contact Us