Skip to content
Work with 3FOLD

AI visibility

What is an AI-visibility audit, and what should it check?

By Adam Collins, Founder, 3Fold Labs

"What ChatGPT, Claude and Grok say about you." over a magnifying glass on a dark teal surface, with three sample assistant answers about Clearflow Plumbing: one correct, one naming no plumbers, and one with an old phone number flagged.

Short answer

An AI-visibility audit checks whether ChatGPT, Claude and Grok, plus Google's AI features for local searches, can find your business, read your website, and describe you accurately. A good one tests crawler access, indexing, listings, page content, and real answers, then hands you a dated snapshot and a ranked list of fixes. It can't guarantee placement.

What does an AI-visibility audit actually measure?

Two things: whether you can show up, and whether you do.

Can you show up? Do you? That's the audit. There's no industry standard for what an audit includes.
The "can" half is technical checks like crawler access and indexing; the "do" half is watching real answers to your customers' questions.
There's no industry standard for what an audit includes.

The "can" part is technical and mostly yes-or-no. Can the AI search crawlers reach your pages? Are your pages indexed? Do your listings exist and match? The "do" part is observation: when someone asks the questions your customers ask, are you named, is what's said about you correct, and which sources did the tool use?

There's no industry standard for what an audit includes. Vendors use "AI visibility audit," "AEO audit," and "GEO audit" for overlapping things (see SEO vs AEO vs GEO). So judge any audit by what it checks, not what it's called.

How is it different from a regular SEO audit?

A lot of it is the same, and that's fine. Google says SEO best practices "remain relevant" for AI Overviews and AI Mode, and that there are "no additional requirements" to appear in them.

What an AI-visibility audit adds:

  • Checks for each AI tool's own crawler, not just Googlebot
  • Bing coverage, because OpenAI names Microsoft among ChatGPT's search partners
  • Live answer testing across several AI tools
  • Accuracy: whether the AI's description of you is right, not just whether you appear

What should it check?

Here's the checklist I'd expect, and what each item is based on.

Audit checklist: crawler access, page controls, indexing, listings, content, structured data, trust facts, live answers and measurement.
Crawler access through measurement, each line shortened from the post's table, which also gives the reason each check is there.
Here's the checklist I'd expect.

Swipe sideways to see every column.

AreaWhat to checkWhy it's on the list
Crawler accessrobots.txt allows OAI-SearchBot, Claude-SearchBot, Claude-User, Googlebot and Bingbot; your host, CDN, or firewall allows OpenAI's published IP ranges; xAI documents no Grok crawlerOpenAI says sites opted out of OAI-SearchBot "will not be shown in ChatGPT search answers." Anthropic says blocking Claude-SearchBot "may reduce your site's visibility" in search results. Google lists allowing crawling "by any CDN or hosting infrastructure."
Page controlsNo stray noindex, nosnippet, NOARCHIVE, or NOCACHE on key pagesGoogle says nosnippet, max-snippet, and noindex are how you limit what Search shows from your pages, AI features included. Bing says NOARCHIVE keeps content out of Copilot responses.
IndexingKey pages indexed in Google (Search Console) and Bing (Bing Webmaster Tools)Google says a page must be "indexed and eligible to be shown in Google Search with a snippet."
ListingsGoogle Business Profile and Bing Places complete, accurate, and matching your siteGoogle lists an up-to-date Business Profile among its AI-feature basics. Bing says Bing Places keeps details "eligible for inclusion in AI-generated responses."
ContentEach service and each town you serve described in plain textGoogle: make sure "important content is available in textual form."
Structured dataPresent, accurate, and matching the visible pageGoogle says structured data should match visible text, and that no special schema is needed for AI features.
Trust factsLicense number matches the state record (in Florida, DBPR), reviews gathered within platform rulesThese are the facts a homeowner checks, and the ones an AI tool can only repeat if they're published.
Live answersA written set of prompts run on ChatGPT, Claude and Grok, with Google as the local check, in private sessions, from your service area, on dated runsOpenAI says location and saved memories shape how ChatGPT rewrites a search.
MeasurementSearch Console, Bing AI Performance report, referral visits in your analyticsGoogle and Bing document their reports (see below).

Checked 10/20/2026. Details change; check the source.

Is it checking the right crawlers?

This is where a lot of checklists go wrong. There are two kinds of AI crawlers, and they do different jobs:

  • Search crawlers decide whether you can appear in answers: OAI-SearchBot (ChatGPT), Claude-SearchBot and Claude-User (Claude), and Googlebot (Google's AI features). xAI doesn't document a crawler for Grok (more on where ChatGPT, Claude and Grok get local business info).
  • Training crawlers collect content to train models: GPTBot (OpenAI), ClaudeBot (Anthropic) and Google-Extended (which Google describes as covering AI training and grounding "in some of Google's other systems").
Two lists: search crawlers OAI-SearchBot, Claude-SearchBot, Claude-User and Googlebot; training crawlers GPTBot, ClaudeBot and Google-Extended.
An audit that marks you down for blocking GPTBot, ClaudeBot or Google-Extended is checking the wrong thing.
There are two kinds of AI crawlers, and they do different jobs.

Getting found on Google and in AI answers

OpenAI documents GPTBot and OAI-SearchBot as independent. Blocking GPTBot is a business choice, not a visibility problem. An audit that marks you down for blocking training crawlers is checking the wrong thing. Same for llms.txt: Google says you don't need "new machine readable files, AI text files, or markup" for its AI features, so a missing llms.txt shouldn't count as a failure (more on whether llms.txt helps with ChatGPT).

How should the AI answers be tested?

Carefully, because a single test means very little. OpenAI says ChatGPT uses an approximate location from your IP address and, if memory is on, "may use relevant saved memories when rewriting a search query." Answers also change from run to run (more on how to get ChatGPT to recommend your business).

A fair test:

  • Write the prompts down first, in your customers' words ("Who's a good HVAC company in Ormond Beach?").
  • Run them in private sessions, with memory off, with search turned on.
  • Run them from, or naming, the towns you actually serve.
  • Repeat them on different dates and log each run.
  • Record who was named, whether the details about you were right, and which sources were cited.
Before and after: one test shifts with each run, location and memory; a fair test uses written prompts, private sessions, your towns and dated logs.
Write the prompts first, run them in private sessions in the towns you serve, repeat on different dates, and record every source.
OpenAI says location and saved memories shape how ChatGPT rewrites a search.

What free tools show AI traffic today?

You can check a lot of this yourself before paying anyone:

  • Google Search Console. Google says AI Overviews and AI Mode appearances are counted in the Performance report under the "Web" search type. They aren't broken out separately.
  • Bing Webmaster Tools. Its AI Performance report shows citations and grounding queries across Microsoft Copilot and partner experiences.

What should you get at the end?

A dated snapshot of what was found, and a list of fixes in order of impact, with the reason for each. Not just a score. A number without the "why" doesn't tell you what to do on Monday.

What an audit can't give you: a promise. OpenAI says placement in ChatGPT search "is not guaranteed," and Google says indexing and serving aren't guaranteed either (more on why ChatGPT doesn't mention your business). Anyone who promises you'll be the AI's top pick is promising something the platforms won't.

Quote from the post: A number without the why doesn't tell you what to do on Monday. Get a dated snapshot and fixes in order of impact.
A score on its own isn't the deliverable; the reasons behind each fix are.
A dated snapshot of what was found, and a list of fixes in order of impact, with the reason for each.

How 3Fold Labs helps

I offer technical and content visibility audits as part of my search visibility work, starting with your core pages, search access and business information. They cover the kinds of checks in this post. From there I fix what you control and use the search data available to judge progress. No audit can guarantee rankings or AI mentions, mine included.

Here's my Getting found on Google and in AI answers.

FAQ

Can I do an AI-visibility audit myself?

Much of it, yes. Check your robots.txt, set up Search Console and Bing Webmaster Tools, compare your listings, and run a written prompt set in private sessions. The harder parts are firewall settings and reading what the cited sources have in common.

How often should it be repeated?

There's no documented right interval. Rerun the live answer tests on a regular schedule you can stick to, and rerun the technical checks any time your site, host, or listings change.

Is a low "AI visibility score" from a tool something to worry about?

Look at how it's calculated. Scores like that aren't a standard from OpenAI, Anthropic, xAI or Google. The checks behind the score matter more than the number.

Will fixing everything on the list get me recommended?

It makes you eligible and easier to verify. It doesn't guarantee a mention. That's true for every audit, no matter who runs it.

Sources

Analytics and session replay stay off until you accept. Form answers are not recorded by these tools. Privacy policy