How to Do an AI Visibility Audit (Step by Step)

Short answer: An AI visibility audit checks what ChatGPT, Gemini, Claude and Perplexity say about your business, then finds why. Ask each assistant 10 to 20 real customer questions, record who gets named, then check crawler access, page content, structured data, business listings and the sources the assistants cite. Fix the biggest gap first.

This guide is for owners and marketers who want a one-off diagnosis they can do in an afternoon, with free tools. It covers what to test, how to run each check, how to score what you find and what to fix first.

What is an AI visibility audit?

An AI visibility audit is a one-time diagnosis of whether AI assistants mention, describe and cite your business correctly, and of the specific causes when they do not. It differs from ongoing tracking: tracking tells you a number each week, while an audit explains the number and produces a fix list.

The audit has two halves. The outside-in half asks the assistants questions and records their answers. The inside-out half inspects the things those answers are built from: whether AI crawlers can reach your pages, whether the pages state the facts plainly, and whether the rest of the web agrees with them. If you want a repeatable scorecard after the audit, see how to measure AI visibility.

What should an AI visibility audit check?

A complete audit checks six areas, from the answers themselves down to the technical access underneath them. The table lists each area, the question it answers and a free way to check it.

How do you run an AI visibility audit step by step?

You run an AI visibility audit in eight steps, starting with the questions your customers actually ask and ending with a ranked fix list. Keep a spreadsheet open; each step adds columns to it.

1. Write 10 to 20 customer questions. Mix category questions ("best emergency electrician in Denver"), comparison questions ("X vs Y for a small office") and brand questions ("is X any good"). Use the words customers use on calls and in reviews, not your internal jargon. Our guide to keyword research for AI search covers how to find them.
2. Ask every assistant, more than once. Put each question to ChatGPT, Gemini, Claude and Perplexity in a fresh chat. Record whether you are named, in what position, whether the facts are right, and which competitors and sources appear. Run each question two or three times, because answers vary between runs.
3. Check crawler access. Open yoursite.com/robots.txt and compare it with each vendor's documentation. OpenAI's crawler overview says sites that opt out of OAI-SearchBot will not be shown in ChatGPT search answers, though they can still appear as navigational links. Perplexity's crawler page recommends allowing PerplexityBot to appear in its results, and Anthropic says blocking Claude-SearchBot may reduce your visibility in Claude's search results. Check your CDN or firewall too, since a bot-blocking rule can refuse these crawlers even when robots.txt allows them.
4. Separate training from search. GPTBot, ClaudeBot and Google-Extended control whether content may be used for model training, and blocking them is a legitimate choice. Make sure nobody blocked the search crawlers by accident while blocking the training ones. Our AI crawlers and robots.txt guide has the full list.
5. Confirm the content is readable. View the page source of your home, services and pricing pages and search for the facts a customer needs: services, area served, prices or price ranges, hours. If they only appear after JavaScript runs, some crawlers may miss them; see JavaScript rendering and AI crawlers.
6. Test structured data. Run key pages through Google's Rich Results Test. Google's LocalBusiness documentation says this markup lets you tell Google about hours, departments, reviews and more. Missing or broken markup is a quick fix; our schema markup guide shows the fields that matter.
7. Check entity consistency. Compare your name, address, phone, website and categories across Google Business Profile, Bing Places, Apple Business Connect, Yelp, industry directories and social profiles. Every mismatch is a reason for an assistant to hesitate or repeat stale facts.
8. Audit the sources assistants cite. Open every citation from step two. List the review sites, directories, "best of" articles and forums that keep appearing, then note which ones mention you and which only mention competitors. That list becomes your off-site plan.

How do you score the results and decide what to fix first?

Score each finding by how many questions it affects and how hard it is to fix, then work from high impact and low effort downwards. Most audits produce the same handful of patterns:

For wrong facts specifically, our guide to fixing wrong AI information walks through where corrections need to land. For the competitor pattern, see why AI recommends your competitor.

Does Google need anything special for AI Overviews?

No, Google says there are no additional requirements to appear in AI Overviews or AI Mode beyond normal Search eligibility. In AI Features and Your Website, Google says a page must be indexed and eligible to be shown with a snippet to be a supporting link, and that meeting the requirements does not guarantee being shown. So the Google part of your audit is ordinary indexing hygiene: no accidental noindex, no nosnippet on pages you want quoted, and pages that answer the question. Check for those controls on your key pages while you are in the page source.

How often should you repeat an AI visibility audit?

Repeat a full audit every quarter, or after any site redesign, migration, CDN change or rebrand, and track the answers more often in between. Redesigns and new firewall rules are the most common way a site that was visible quietly stops being fetched.

Between audits, re-ask your question list monthly at minimum. Expect some movement even when nothing changed, since assistants vary their answers; look for trends across several runs rather than reacting to one.

Can you automate an AI visibility audit?

Yes, partly: trackers can automate the asking, and agents can automate some of the fixing, but someone still has to approve what changes on your site. Monitoring tools run your question list on a schedule and log mentions and citations; our best AEO tools of 2026 compares them.

Alphaa is an AI agent that does both halves for one business. It asks ChatGPT, Gemini, Claude and Perplexity your customers' questions every week, runs a 23-point check of what those assistants can read on your site, and drafts the fixes, such as FAQ pages, structured data and llms.txt, for you to approve before anything is published to WordPress, Shopify or Webflow. Plans start at $99 a month. To see the outside-in half for free, run the 60-second AI check.

What else do people ask about AI visibility audits?

How long does an AI visibility audit take?

A manual audit for one business takes about an afternoon with 10 to 20 questions. Most of the time goes on asking the assistants and opening their citations; the technical checks take under an hour once you know where to look.

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

No, though they overlap. An SEO audit checks how Google ranks your pages, while an AI visibility audit also tests what assistants actually say, which crawlers they use and which third-party sources they trust for your category.

Should I block GPTBot during an audit?

That is a business decision, not an audit finding. OpenAI says GPTBot is used for training and OAI-SearchBot for search, and the settings are independent, so you can block training while still allowing ChatGPT search.

Why do I get different answers when I repeat a question?

AI assistants generate each answer fresh, so wording, order and even which businesses appear can change between runs. That is why an audit asks each question more than once and looks at the pattern.

Can an audit guarantee that ChatGPT will recommend me?

No. An audit finds and removes the obstacles you control, such as blocked crawlers, missing facts and inconsistent listings, but no one can force an assistant to name a business.