AI Visibility Audit: What It Checks and What It Cannot

View this page

What an AI visibility audit checks, how to run the same checks by hand, why answers differ between runs, and what the published evidence says changes them.

Organisation
WavX Solutions
Telephone
+919310079927

Description

Service AI visibility audit: does ChatGPT know your business?

An AI visibility audit checks three things: whether the search crawlers behind ChatGPT, Perplexity, Claude, Google and Bing can reach your pages, what those tools say when buyers ask about your category, and what exists about you on other sites for them to draw on. It ends in a written list of fixes. It cannot promise that any tool will name you.

Discuss your SEO/GEO/AEO strategy Last updated 2 October 2026

What the audit answers

A buyer asks ChatGPT, Perplexity, Gemini or Google for "the best accountant for a small exporter in Gurgaon" and gets a short list. The audit answers four questions about that moment:

Can the engines reach you? Each one uses a search crawler, and a blocked or unreadable site cannot be used as a source.

Do they name you? Measured by running the questions your buyers ask, many times, and counting.

What are they reading when they answer? The pages each tool cites for your category, and whether you are on them.

What would change the result, in what order? A list of fixes, each marked with the evidence behind it.

It is the first step of WavX's generative engine optimisation service , and it is also sold on its own.

What is checked

Area

What WavX checks

Why it matters

Search crawler access

robots.txt , CDN and firewall rules for Googlebot, Bing's crawler, OAI-SearchBot, PerplexityBot and Claude-SearchBot

OpenAI states that sites opted out of OAI-SearchBot "will not be shown in ChatGPT search answers". Perplexity recommends allowing PerplexityBot in order to appear in its results, and Anthropic says that blocking Claude-SearchBot may reduce visibility in Claude's search results

Training crawlers

Rules for GPTBot, ClaudeBot and the Google-Extended token

These are separate decisions. Google states that Google-Extended "does not impact a site's inclusion in Google Search". Blocking a training crawler is a choice, and it should be made on purpose

What the server returns

Whether the text of each key page is in the HTML, or appears only after JavaScript runs

Google says it can process JavaScript but that such sites are more complex to get right. The OpenAI, Perplexity and Anthropic crawler pages do not say whether their crawlers run JavaScript, so text in the HTML is the safe assumption

Indexing and eligibility

Index status in Google and Bing; noindex , nosnippet and max-snippet rules; the Search generative AI setting in Search Console

Google requires a page to be indexed and eligible for a snippet. Since 2026 a site must also be included under Search Console's generative AI control, which is on by default and can be switched off

First-party pages

Whether your pages state plainly what you do, where, for whom and at what price, with dates and sources

An engine can only repeat what a page says clearly

Third-party presence

Listings, review profiles, Google Business Profile, Bing Places, directories and "best of" lists for your category; whether name, address and phone agree

See the evidence table below. This is usually where the gap is

Prompt runs

A fixed set of buyer prompts, each run repeatedly across ChatGPT, Perplexity, Gemini and Claude, with named competitors

A count over many runs is a measurement. One run is an anecdote

Measurement baseline

Search Console's generative AI report, Bing Webmaster Tools' AI Performance report, AI referrals in analytics

So that later changes can be compared with a starting point

The crawler names and what each one is for are listed in the AI crawler directory.

How to check your own visibility by hand

None of this needs a paid tool. Allow two to three hours.

Write ten prompts in a buyer's words. Three that name your category and place ("best physiotherapy clinic in Sector 56 Gurgaon"), three that describe a problem ("my Shopify orders are not reaching Tally, who can fix it"), two that compare you with a named competitor, and two that name you ("what is [your business]", "is [your business] reliable").

Open ChatGPT, Perplexity, Gemini and Claude. Start a new conversation for every run, so that an earlier answer does not shape the next one. Where a tool offers web search as an option, ask it to search.

Run each prompt at least five times per tool. Five is the least that shows the variation. The SparkToro authors suggest dozens to hundreds of runs before treating the figure as a metric.

Record four things per run: did your business appear, how was it described, which competitors appeared, and which sources were linked.

Count the share of runs in which you appear. That percentage is the number to keep. Ignore the order of the list.

Open the sources the tools linked. They are the pages the answer was built from: directories, review sites, comparison articles, sometimes a competitor's own page. Note the ones your business is absent from.

Read your robots.txt at yourdomain.com/robots.txt and look for Disallow rules against the search crawlers in the table above.

Open Search Console and Bing Webmaster Tools. In Search Console, look for the generative AI performance report and check under Settings that Search generative AI is set to include your site. In Bing Webmaster Tools, open AI Performance.

If step 5 gives zero across every tool, the troubleshooting guide lists the causes in order of likelihood.

Why two runs give two answers

Generated answers are written afresh each time, and the engines say so. Google's documentation notes that AI Overviews and AI Mode "may use different models and techniques, so the set of responses and links they show will vary", and that AI Overviews often do not appear at all.

In January 2026 SparkToro and Gumshoe had 600 volunteers run 12 prompts through ChatGPT, Claude and Google's AI features 2,961 times. For ChatGPT and Google's AI, the chance of the same list of brands appearing in any two of a hundred responses was under 1 in 100, and Claude was only slightly more consistent. The same list in the same order came up roughly once in 1,000 runs. The authors, who describe themselves as not being professional researchers, concluded that position in the list is meaningless and that the share of runs in which a brand appears is a reasonable measure.

The audit therefore reports frequencies, with the date, the tool and the number of runs beside every figure.

What the evidence says changes visibility

This field is full of confident advice. The table sets out what has been published, and how strong each piece is.

Claim

What the evidence shows

Strength

"Google's AI features need special optimisation"

Google: "There are no additional requirements to appear in AI Overviews or AI Mode, nor other special optimizations necessary." The page must be indexed and eligible for a snippet

Vendor documentation

"Mentions on other sites matter"

Across 75,000 brands, branded web mentions correlated 0.664 with being mentioned in Google AI Overviews. Number of pages on the brand's own site: 0.17

Ahrefs, May 2025. Large and observational: correlation only, established brands only, Google AI Overviews only

"Add schema to get cited"

1,885 pages that added JSON-LD, against 4,000 matched controls: no meaningful change in AI Mode (+2.4%) or ChatGPT (+2.2%), and a small decline in AI Overviews (−4.6%)

Ahrefs, May 2026. Matched control, but only pages that were already cited often, over 30 days

"Publish llms.txt"

Of 137,210 domains, 28% published one, and 97% of those files received no requests at all in May 2026. Google states that Search ignores the file

Ahrefs log study, June 2026; vendor documentation

"Sourced facts get quoted"

Adding citations, quotations and statistics gave a 30–40% relative improvement on the study's main visibility metric. Keyword stuffing gave little to none

Princeton University and IIT Delhi researchers, KDD 2024. Controlled, but in a simulated engine fed the top five Google results, so it describes pages already retrieved

"Buy mentions and listings in bulk"

Google: "seeking inauthentic 'mentions' across the web isn't as helpful as it might seem", and its spam systems apply to AI features

Two conclusions follow. What other sites say about a business is associated with AI visibility far more strongly than anything on the business's own site. And the on-site work that has support is plain: be reachable, be indexed, and state sourced facts clearly. How AI answer engines choose sources goes through the mechanics.

What you receive

A written report, in plain language, in six parts.

Part

Contents

1. Access

Each search and training crawler, whether it is allowed, and where any block sits ( robots.txt , CDN, firewall)

2. Index and eligibility

Pages indexed in Google and Bing, pages excluded and why, snippet rules, the Search Console generative AI setting

3. Prompt results

Every prompt, every engine, the number of runs, the share of runs in which you and each named competitor appeared, and how you were described

4. Sources cited

The third-party pages the engines linked for your category, and whether you are on each

5. First-party gaps

Pages that fail to state a fact a buyer asks about, or state it without a date or source

6. Fix list

Each fix marked urgent, important or optional, with the evidence for it and who would need to do it

Part 4 is often the most useful, because it names the actual directories, lists and review sites to work on.

What the audit cannot do

It cannot promise a mention. No vendor controls what an engine generates. Google's own hiring guidance says no one can guarantee a first-place ranking, and the same holds for AI answers.

It cannot show what your buyers typed. No engine publishes the questions people ask it. The nearest thing is Bing's report of a sample of the phrases its system used for retrieval. The prompt set is an informed sample.

It cannot see inside the engines. Google states that no third-party tool has access to its internal ranking or AI systems. The audit measures outputs.

Its numbers age. Models and indexes change. A prompt result is true for the date it was run.

When you should not buy one

Your site is not in Google's index. Fix that first; the guide to a website that is not showing on Google covers it. AI visibility work on an unindexed site is premature.

You already know the answer. If the business has no listings, no reviews and no mentions anywhere, the hand check above will show zero and the cause is not in doubt. Spend the money on getting listed and reviewed.

You sell to people who do not ask AI tools. If your enquiries come from referrals and walk-ins, a technical SEO audit and a complete Google Business Profile will do more.

An audit is worth paying for when the hand check gives a mixed picture, when competitors appear and you do not, or when a developer is needed to find out why a crawler cannot read the site. Follow-up measurement is described under AI citation and referral tracking .

Frequently asked questions

Can I check my own AI visibility without paying anyone?

Yes. Write ten prompts a buyer would type, run each one several times in a fresh conversation in ChatGPT, Perplexity, Gemini and Claude, and count how often your business appears and which sources are cited. The steps are on this page. An audit adds the crawler, indexing and off-site checks and turns the result into an ordered list of fixes.

Why does ChatGPT mention us one day and not the next?

Because the answers are generated afresh each time. In a January 2026 experiment by SparkToro and Gumshoe, a prompt asked of ChatGPT a hundred times had a less than 1 in 100 chance of returning the same list of brands twice. One answer tells you very little. The share of runs in which you appear, counted over many runs, is the number worth tracking.

Will the audit get us mentioned in ChatGPT or Google AI Overviews?

No. An audit finds what is blocking or weakening your visibility and says what to fix. Whether an engine then names you is decided by that engine, and no company outside it controls the outcome. We describe the work and measure what happens; we do not promise a mention, a citation or a placement.

How much does an AI visibility audit cost?

It is quoted after a short call, because the size of the job depends on how many prompts, engines and competitors are covered and how large the site is. WavX does not publish a price for GEO work. Any work that follows is quoted after the audit, from its findings.

Do we need this if we already do SEO?

Much of it overlaps. Google states that its AI features draw on the same index and ranking systems as ordinary Search, so sound SEO is most of the Google side. The audit adds the checks SEO reports usually skip: the non-Google search crawlers, repeated prompt runs, and the third-party pages the engines cite.

Sources

Google Search Central: AI features and your website · read 2 October 2026

Google Search Central: Optimizing your website for generative AI features on Google Search · read 2 October 2026

Search Console Help: Search generative AI control · read 2 October 2026

Search Console Help: Generative AI performance report (Search) · read 2 October 2026

Google: common crawlers, including the Google-Extended token · read 2 October 2026

Google Search Central: Do you need an SEO? · read 2 October 2026

OpenAI: overview of OpenAI crawlers · read 2 October 2026

Perplexity: Perplexity crawlers · read 2 October 2026

Anthropic: does Anthropic crawl data from the web, and how can site owners block the crawler? · read 2 October 2026

Bing Webmaster Blog: Introducing AI Performance in Bing Webmaster Tools (10 February 2026) · read 2 October 2026

Ahrefs: An analysis of AI Overview brand visibility factors, 75K brands (26 May 2025) · read 2 October 2026

Ahrefs: We tracked 1,885 pages adding schema (11 May 2026) · read 2 October 2026

Ahrefs: We analyzed 137K sites, 97% of llms.txt files never get read (15 June 2026) · read 2 October 2026

Aggarwal et al., GEO: Generative Engine Optimization (KDD 2024), arXiv 2311.09735 · read 2 October 2026

SparkToro: AIs are highly inconsistent when recommending brands or products (28 January 2026) · read 2 October 2026

Related

Generative engine optimisation (GEO) services

AI citation and referral tracking

Answer engine optimisation (AEO) services

Technical SEO audit

Build your own software — your way, your pricing.

WavX Solutions is here to create your own software in a fully custom way, built exactly how you work — with a pricing model that fits your business. Connect now and let's build it.

Contact Now helpwavx@gmail.com