Skip to main content Scroll Top

AI Visibility Audit: Find Out Why AI Does Not Mention You

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

An AI visibility audit tells you why AI assistants do not mention your business: a tracker tells you that you are absent, an audit tells you why. Six stages, run in order, from server logs to a dated baseline — and most sites fail at the first two, where the fix costs hours rather than quarters.

On this page · 11 sections
  1. What is an AI visibility audit?
  2. What does the audit actually check?
  3. What do these audits usually find?
  4. What do you receive at the end?
  5. How is this different from buying a visibility tool?
  6. What should you ask before commissioning one?
  7. What an audit will not do
  8. How long does it take and how is it scoped?
  9. Who actually needs an audit, and who does not?
  10. What happens after the audit?
  11. More from Progression Agency

The short answerAn AI visibility audit is a fixed-scope diagnostic covering six stages in order: can AI crawlers reach your pages, does the answer exist in the HTML without JavaScript, are the pages indexed, does each section answer its own heading, does the site describe the business consistently, and what do the assistants actually say today. It runs in one to two weeks and returns six inspectable deliverables, including a prompt set and baseline you keep permanently.

Updated September 2026. Crawler names and assistant behaviour change; the technical checks here are re-verified quarterly.

What an AI visibility audit checks, in orderWhat an AI visibility audit checks, in order
Most audits we run fail at stage one or two, where the fix costs hours rather than quarters.

What is an AI visibility audit?

A fixed-scope diagnostic that establishes why a business does or does not appear in AI-generated answers. It checks six things in order — crawler access, rendering, indexation, answer structure, entity consistency, and a measured baseline — and returns findings sequenced by expected effect rather than by issue count.

The distinction from a visibility subscription matters and is the reason this page exists. A tracking tool tells you where you stand this month. It cannot tell you why, because the reasons live in places a tracker does not look: your server logs, your rendered HTML, your markup and your page structure. Businesses routinely buy the measurement, confirm the problem monthly for a year, and never diagnose it.

An audit is the other purchase. It is bounded, it happens once, and it produces a list of specific changes in the order they should be made. Most of what it finds is unglamorous and cheap to fix, which is precisely why it is worth finding.

Access — Stage 1. Can the crawlers fetch you at all.
Render — Stage 2. Does the answer exist without JavaScript.
Index — Stage 3. Are the pages in the underlying web index.
Structure — Stage 4. Does the answer lead its section.
Entity — Stage 5. Do you describe yourself consistently.
Baseline — Stage 6. What do the assistants say today.

What does the audit actually check?

Six stages, run in order, because a failure at stage one makes every finding downstream irrelevant. There is no point restructuring content on a site the crawlers cannot reach.

Stage 1 — crawler access, proved from logs

We pull ninety days of server logs and count actual fetches by GPTBot, OAI-SearchBot, ChatGPT-User, PerplexityBot, Google-Extended, ClaudeBot and Bingbot. A robots.txt file states an intention; a log records what happened. We also request your pages with each crawler’s user-agent string, because a CDN or WAF can block a bot that robots.txt permits, and it does so invisibly.

Stage 2 — rendering without JavaScript

We fetch every commercial template with JavaScript disabled and read what comes back. This is where the expensive failures live: pages that look perfect in a browser and return a few hundred bytes of scaffolding to anything that does not execute scripts. We test templates rather than pages, because failures cluster by template.

Stage 3 — indexation

The AI search surfaces draw on web indexes, so pages that are not indexed cannot be retrieved from them. We check what is indexed against what should be, and look for anything actively suppressing it — a stray noindex, a conflicting canonical, a redirect chain.

Stage 4 — answer structure

Retrieval selects passages rather than pages. We check whether each section answers its own heading immediately, whether the answer is complete within the first few hundred characters, and whether the passage still makes sense lifted out of context.

Stage 5 — entity consistency

We collect every description of the business — page copy, Organization schema, directory listings, social profiles — and identify where they disagree. Inconsistency here is the commonest cause of an assistant describing your business incorrectly.

Stage 6 — baseline measurement

We write a fixed prompt set covering your real commercial questions, run it, and record what the assistants say today, including which competitors appear. Without this, no later claim of improvement can be checked by anyone, including us.

Server logs — Evidence. The only proof of crawler access.
JavaScript-off fetch — Evidence. What the crawler actually reads.
Bot-agent requests — Evidence. Status codes served to each crawler.
Schema validator — Evidence. Markup against the visible copy.
Rendered heading tree — Evidence. Whether structure is real or styling.
Fixed prompt set — Evidence. The baseline everything is measured against.
What we typically findWhat we typically find
The bad rows are what an audit exists to find. Each one is specific and fixable.

What do these audits usually find?

Content that only exists after JavaScript runs, and no baseline of any kind. Those two appear in the large majority of audits we run, and both are invisible to every browser-based check a marketing team would normally perform.

The crawler blocks are less common but more decisive when present. A blanket disallow left over from a staging environment, or a bot-management rule in a CDN that nobody on the marketing side knows exists, takes a site from competing badly to not competing at all. It is also the cheapest thing on this page to fix — usually minutes.

Empty without JavaScript — Finding. The expensive failure.
Crawler blocked — Finding. The cheapest fix.
403 to bots, 200 to browsers — Finding. Invisible in every browser test.
Answer buried — Finding. The passage is never selected.
Schema contradicts the page — Finding. Worse than no markup.
Three self-descriptions — Finding. You do not choose which is repeated.
Where audits actually fail, across the sites we have auditedWhere audits actually fail, across the sites we have audited
Directional, from our own audit work rather than a published study.
What we find, what it costs to fix, and how fast it shows
FindingFrequencyCost to fixTime to show in logs or answers
Content missing without JavaScriptVery commonConfiguration, sometimes a build changeWeeks
No baseline existsVery commonHalf a dayImmediate, as a baseline
Entity description inconsistentCommonAn afternoonWeeks to months
Answer buried in its sectionCommonEditorial pass per pageWeeks
Schema contradicts the pageCommonHoursWeeks
Crawler blocked in robots or WAFOccasionalMinutesDays
Bots served a different status codeRareHoursDays
How findings get prioritisedHow findings get prioritised
The report is ordered by this, not by issue count. A three-hundred-item list is a data export, not an audit.

What do you receive at the end?

Six deliverables, every one of them inspectable, and none of them dependent on us. The prompt set in particular is yours from the first day — a measurement instrument you cannot run yourself is a dependency, not a measurement.

What you receiveWhat you receive
All six are yours to keep and to hand to any other provider.
Crawler access report — Deliverable. Which bots arrived, and which URLs.
Render audit — Deliverable. Per template, with output shown.
Structure and schema findings — Deliverable. Answer position and markup conflicts.
Entity file — Deliverable. One description, applied everywhere.
Prompt set and baseline — Deliverable. Yours permanently.
Remediation plan — Deliverable. Sequenced by expected effect.
The six deliverables
DeliverableWhat it containsWhat you can do with it
Crawler access reportFetches by bot and by URL across 90 daysProve or disprove access in one page
Render auditEach template’s output with JavaScript disabledHand straight to a developer
Structure and schema findingsAnswer position, heading tree, markup conflictsBrief an editor and a developer
Entity fileOne canonical description and every place to apply itFix in an afternoon
Prompt set and baselineFixed wording, dated results, competitors namedRe-run monthly yourself, forever
Remediation planSequenced by expected effect, with dependenciesDecide what to fund

How is this different from buying a visibility tool?

A tool measures. An audit diagnoses. Both are legitimate purchases and they answer different questions — which is why the honest sequence is usually audit first, subscription afterwards.

If you already know you are absent from AI answers, a monthly subscription will confirm it every month at a monthly cost. What you need first is the reason, and the reason is in your logs and your rendered HTML. Once the causes are fixed, a tracker earns its fee, because then there is something to detect.

Audit against subscriptionAudit against subscription
1 = no, 3 = yes. A tracker is an instrument; an audit is a diagnosis.
Audit or subscription
QuestionFixed-scope auditVisibility subscription
Where do I stand?Yes, onceYes, continuously
Why am I not appearing?YesNo
Can crawlers reach my site?YesNo
Does my page render without JavaScript?YesNo
What should I change, in what order?YesNo
Did last month’s change work?Only if re-runYes
Cost shapeOne project feeRecurring

Want the diagnosis?

A fixed-scope AI visibility audit: crawler access from your logs, render testing per template, structure and entity findings, and a prompt set with a dated baseline. One to two weeks. You keep everything, including the prompt set.

/contact-us

Do you want our server logs? — Ask. The question that separates doers from describers.
Which crawlers, by name? — Ask. The list is short and public.
Will you render without JavaScript? — Ask. Ten minutes per template.
What is measured before you start? — Ask. Without a baseline, nothing is checkable.
Who implements the fixes? — Ask. Changes your cost and timeline.
What will you not claim? — Ask. Honest answers name limits.

What should you ask before commissioning one?

Whether they want your server logs. Everything else in an AI visibility pitch is assertion; the logs are the one artefact that settles whether anyone has looked at your site.

Do you want our server logs?

If the answer is no, the crawler-access stage is guesswork, and that is the stage most sites actually fail.

Which crawler user agents will you test, by name?

The list is short and public. A vague answer about AI bots means nobody has opened a log.

Will you render our templates without JavaScript and show us?

Ten minutes per template, and it finds the most expensive failure there is.

What will you measure before starting?

A baseline taken against fixed wording, handed to us. Without it, no later claim is checkable.

Who implements the findings?

Much of this is developer work. A provider producing recommendations for your team to action should say so, because it changes your cost and your timeline.

What will you not claim?

Guaranteed placement in any AI answer cannot be sold by anyone. An honest scope names its limits.

Issue-count reports — Not this. A crawler export is not an audit.
Proprietary scores — Not this. Comparable to nothing outside the tool.
Guaranteed placements — Not this. Nobody can sell them.
Audit without logs — Not this. The one checkable artefact, skipped.
Recommendations with no order — Not this. Sequence is most of the value.
Findings you cannot verify — Not this. Every claim should be traceable.

What an audit will not do

It will not guarantee placement in any AI answer, it will not produce a proprietary score, and it will not make the changes for you unless that is scoped separately.

It also will not tell you that everything is fine when it is not, which is occasionally what a client is hoping for. The most common uncomfortable finding is that the content programme running for the last two years was invisible to the systems it was aimed at, because the pages it produced never rendered for them.

In scope, out of scope
In scopeOut of scope
Crawler accessAnalysed from your logsChanging your CDN configuration
RenderingTested per templateImplementing server-side rendering
StructureReviewed and specifiedRewriting every page
EntityOne description producedUpdating third-party listings
MeasurementPrompt set and baseline builtOngoing monthly measurement
OutcomesSequenced planGuaranteed placement, which nobody can sell
What the audit runs toWhat the audit runs to
The critical path is nearly always how quickly we get server logs.

Want this done for your site?We run the six-stage audit from your server logs to a dated baseline and hand you every finding as an inspectable deliverable.

Get a free proposal

How long does it take and how is it scoped?

One to two weeks from the point we have server logs and staging access. Scope is driven by template count rather than page count, because rendering failures cluster by template — a two-hundred-page site with four templates is less work than a forty-page site with fifteen.

The critical path is almost always log access. On sites where the hosting is managed by a third party, requesting ninety days of raw logs can take longer than the analysis does, so it is worth starting that request before anything else.

Who actually needs an audit, and who does not?

You need one if you suspect you are absent and do not know why. You do not need one if nobody has yet checked whether you are absent at all — that check is free and takes an hour.

You probably need one

If a tracker or a manual check shows you are missing from answers your competitors appear in, and nobody internally can say whether AI crawlers reach your site.

You probably need one

If your site is built on a modern JavaScript framework and nobody has fetched a page with scripts disabled to see what remains.

You probably do not need one yet

If you have never asked the assistants your buyers’ questions. Do that first; it is free, it takes an hour, and it may tell you the problem is smaller than you feared.

You probably do not need one yet

If the constraint is obviously corroboration — nobody outside your business has ever written about you. That is a real problem and a technical audit will not find it.

If the constraint is obviously corroboration — nobody outside your business has ever written about you.

What happens after the audit?

You get a sequenced plan and a choice: action it with your own developers, or have us do it. Both are normal, and they are different engagements priced differently.

What should not happen is the plan sitting unactioned while a subscription measures the unchanged problem every month. If the findings are not going to be implemented, the honest advice is not to commission the audit in the first place.

Implementing it yourself

Entirely reasonable with developer access. The render and crawler fixes are usually configuration changes, and the editorial restructure is a bounded writing project.

Having us implement it

Scoped separately from the audit, so you can see what each costs. The rewrites and template changes are the bulk of it.

Re-measuring

Re-run the prompt set monthly with identical wording. That is the only way anyone — including us — can tell whether the work moved anything.

Re-checking quarterly

Crawler names and behaviour change. A quarterly re-check of access is cheap insurance against silently losing what you fixed.

What the audit producesWhat the audit produces
Nothing here is a number only we can calculate.

More from Progression Agency

The rest of the library, by what you are trying to do

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

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 AI visibility audit?
A fixed-scope diagnostic that establishes why a business does or does not appear in AI-generated answers, covering crawler access, rendering, indexation, answer structure, entity consistency and a measured baseline.
What does an AI visibility audit include?
Six deliverables: a crawler access report from your server logs, a render audit per template, structure and schema findings, an entity file, a prompt set with a dated baseline, and a remediation plan sequenced by expected effect.
How much does an AI visibility audit cost?
It is scoped like a technical audit — driven by template count and site size rather than by a per-page or per-keyword rate.
How long does an AI visibility audit take?
One to two weeks from the point we have server logs and staging access. Log access is usually the critical path, not the analysis.
What is the difference between an AI visibility audit and an AI visibility tool?
A tool measures where you stand each month. An audit diagnoses why. Tools cannot read server logs or test rendering, which is where most causes live.
Do I need to give you server logs?
Yes, and it is the single most important input. Ninety days is ideal. Without logs, crawler access is guesswork, and that is the stage most sites actually fail.
What is the most common finding?
Content that only exists after JavaScript runs. It looks perfect in a browser and returns almost nothing to the crawlers that matter.
What is the cheapest thing you usually find?
A crawler blocked by accident in robots.txt or a CDN rule. It takes minutes to fix and it is the difference between competing badly and not competing at all.
Can you run the audit without a staging environment?
Yes, though render testing is safer and faster with one. We can test against production if necessary.
Do you fix what you find?
Implementation is scoped separately. Some clients action the findings with their own developers; some ask us to. Both are fine, and they are different engagements with different costs.
Will an audit guarantee I appear in ChatGPT?
No, and nobody can sell that. What an audit guarantees is that you will know why you do not, and what to change.
What is an AI SEO audit?
The same thing under a different name. AI SEO audit, AEO audit and AI visibility audit all describe this diagnostic.
What is an AEO audit?
Also the same thing. The label varies by who is speaking; the six stages do not.
What is in an AI visibility report?
The findings and the baseline: which crawlers arrived, what each template renders, where answers sit, where markup conflicts, and what the assistants currently say about you and your competitors.
Do you provide a visibility score?
No. Proprietary composite scores are not comparable to anything outside the tool that produced them. We report mentions, citations, competitors and the technical findings directly.
How many prompts are in the baseline?
Sixty to two hundred, covering your real commercial questions. Fewer than sixty and normal run-to-run variation looks like a trend.
Can I re-run the baseline myself?
Yes — that is the point of handing it over. The prompt set is yours permanently and needs no tool to run.
How often should the audit be repeated?
The full audit rarely needs repeating. Re-check crawler behaviour quarterly, because user agents and rules change, and re-run the prompt set monthly.
Does this replace an SEO audit?
No. It overlaps on indexation and technical health and adds crawler access for a second set of bots, rendering strictness, passage structure and prompt-set measurement.
What if the audit finds nothing wrong?
That is a useful result and it happens. It means the constraint is corroboration or content depth rather than technical, and the remediation plan says so.
Who is this for?
Businesses that already suspect they are invisible in AI answers and want the reason rather than a monthly confirmation of the symptom.
Do small sites need this?
A twelve-page site can often be checked more cheaply and completely than a large one, because the technical surface is small enough to get entirely right.

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

Get a free proposal

Check, checker, score, report or audit?Buyers use these five words interchangeably and they are not the same purchase. A check is something you run yourself in minutes. A checker is a tool that automates it. A score is a single number, usually a bad one to manage by. A report is the written output of a measurement run. An audit is a diagnosis that ends in a scoped plan.

An AI visibility check versus an AI visibility checker

An AI visibility check is an action. You ask your buyers' questions in two or three assistants, note whether you are named, and write down what you find. It costs an hour and no money, and for many sites it is genuinely enough to establish whether there is a problem.

An AI visibility checker is a tool that automates that action. Free AI visibility checkers exist and are worth running, with one caveat worth stating plainly: a single automated run tells you very little, because answers vary between runs of the identical prompt. A checker that reports one result from one run is showing you a snapshot and calling it a measurement.

Neither a check nor a checker diagnoses anything. They tell you that you are absent. They do not tell you whether the cause is a blocked crawler, a page that needs JavaScript to render, an answer buried in paragraph four, or sources that disagree about what your business does.

When a free check is the right purchase

When you have never looked. Run it before commissioning anything, because the result sometimes makes the rest of the conversation unnecessary.

When a checker is worth automating

Past roughly thirty prompts across more than one platform, where doing it by hand stops being realistic.

An AI visibility score is the wrong thing to manage by

A single AI visibility score merges at least four different measurements: whether you are named at all, how often relative to competitors, whether the description is accurate, and which questions you appear on. Each has a different cause and a different fix.

When a blended score falls, it cannot tell you which of the four moved. That is not a minor reporting preference — it is the difference between a number you can act on and a number you can only worry about. We report the four separately for that reason.

What we report instead

Presence, share of voice, sentiment and prompt coverage, each on its own line, with the platform and date attached.

Why scores stay popular anyway

They fit on a slide. That is the whole of the argument for them.

An AI visibility report versus an AI visibility audit

An AI visibility report is the written output of a measurement run: what the prompt set returned, on which platforms, on what date, with the answers stored. It describes a state.

An AI visibility audit explains that state. It reads your server logs to establish whether AI crawlers actually arrived, fetches your templates with JavaScript disabled to see what machines receive, checks whether sections answer their own headings, and compares your structured data against your visible copy. It ends with a scoped, priced plan.

Both are useful and they are not substitutes. A report without an audit tells you that you are absent. An audit without a report has no baseline to measure improvement against, which is why ours includes one.

What an LLM visibility audit covers

The same technical ground, framed around retrieval: whether GPTBot, OAI-SearchBot, ClaudeBot and PerplexityBot reach the page, and whether a clean passage exists for them to lift. LLM visibility audit and AI visibility audit describe the same work.

What an LLM visibility checker cannot do

Tell you why a model selected a competitor. No provider exposes model reasoning, so every explanation offered is inference rather than measurement.

What each of the five actually gets you

You buyYou receiveTimeCostTells you why?
AI visibility checkYour own notes from asking the questions1 hourFreeNo
AI visibility checkerAn automated snapshot, usually one runMinutesFree to lowNo
AI visibility scoreA single blended numberInstantVariesNo
AI visibility reportA measurement run, stored and datedDaysLowPartly
AI visibility auditDiagnosis from logs, rendering and structure, plus a baseline1 week$1,000-$4,000 publishedYes

Frequently asked questions

What is an AI visibility check?

Asking your buyers' questions in two or three assistants and recording whether you are named. It takes about an hour, costs nothing, and it is the right first step before commissioning anything. It establishes that there is a problem; it does not explain the cause.

Is there a free AI visibility checker?

Several exist and they are worth running. The caveat is that answers vary between runs of the identical prompt, so a single automated run is a snapshot rather than a measurement. Run any prompt three to five times before drawing a conclusion from it.

What is an AI visibility score?

A single number merging presence, share of voice, sentiment and prompt coverage. We do not manage by one, because when it falls it cannot tell you which of the four moved, and each has a different fix.

What is the difference between an AI visibility report and an audit?

A report describes a state: what the prompt set returned, on which platforms, on what date. An audit explains it, using server logs, scriptless rendering tests, passage structure and schema comparison, and ends with a scoped plan. Ours includes a report as the baseline.

Is an LLM visibility audit the same as an AI visibility audit?

Yes. The LLM framing emphasises the retrieval layer and the AI framing is the marketing convention, but the technical work is identical: crawler access, rendering, passage structure, entity consistency and a measured baseline.

Can any checker tell me why a model chose a competitor?

No. No provider exposes model reasoning, so every explanation is inference. A tool can tell you that you are absent; only a diagnosis of your own site explains it.

Do I need an audit if a free checker already told me I am invisible?

Only if you want to know why. The checker result and the audit answer different questions, and for a simple site the free technical checks may be enough to find the cause yourself.

Related

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