Skip to content
Back to writing
Method

AI visibility audit: how to run your own in an afternoon

You do not need to pay for an AI visibility audit. Here is how to run one yourself in an afternoon, with a fixed prompt set, four engines, one sheet and a test for real gaps.

11 Nov 2026 - 6 min read
On this pageShow 9Hide

An AI visibility audit answers one question: when buyers ask AI engines about your category, are you in the answer, and if not, who is and why? Some agencies sell this as a standalone deliverable. You can run it yourself in an afternoon with nothing but the engines, a spreadsheet and some discipline about what you record.

This guide is the practical companion to our post on why most AEO audits are a waste of money. That post argues the diagnosis should never be the expensive part. This one shows you how to do the diagnosis yourself, so you can spend the budget on the fix.

What you need

  • One to three topics you sell into, each with the Google keyword a buyer would search for it
  • Access to ChatGPT, Gemini and Perplexity, and a Google search window
  • A spreadsheet with one row per answer
  • About four hours
Free scan

Check your AI visibility free

See which buying questions ChatGPT, Perplexity, Gemini and Google's AI answers name you for, and who they name instead. No call, no card.

Hour one: build the prompt set

The prompt set decides the result, so build it before you look at a single answer. For each topic, write five prompts from different angles:

  1. Category: "best payroll software for restaurants"
  2. Positioning: the thing you claim to do differently, asked as a need
  3. Sector: the category for a specific kind of buyer
  4. Outcome: the result the buyer wants, in their words
  5. Comparison: "alternatives to" the competitor you lose deals to

Write them the way a buyer types, not the way your homepage talks. Keep your brand name out of them. A prompt that names you only tests whether an engine knows you exist; it says nothing about whether it recommends you.

Hour two: run every prompt on every engine

Put each prompt to Google (to see whether an AI Overview appears), ChatGPT, Gemini and Perplexity. Five prompts on four engines is 20 answers per topic.

Use a clean session. OpenAI says that with Reference chat history turned on, ChatGPT can use information from past conversations to personalize future responses (OpenAI). Turn memory off where you can, so you see what a stranger would see.

On Google, expect gaps that are not yours. Google says AI Overviews are shown only when its systems decide they add something to classic Search, and "as such, often don't trigger" (Google Search Central). Record "no AI Overview" as its own result, not as a loss.

For every answer, record:

  • date, engine and prompt
  • every brand named, in order
  • whether you are named, and in what position
  • how you are described, and whether it is accurate
  • the pages cited, where the engine shows them

Once per topic, also record where your site ranks on Google for the keyword behind the prompts.

Hour three: read the sources

The sources column is the most useful part of the audit. When ChatGPT searches the web, its answers may include citations and a Sources panel (OpenAI). Perplexity says each answer includes numbered citations linking to the original sources (Perplexity). Gemini may show a Sources button, though Google notes that not every response has sources (Google). AI Overviews show supporting links.

Sort the cited pages by type: roundups of the best options, comparison pages, review sites, forum threads, competitors' own pages and your own pages. Then take every answer where a competitor was named and you were not, and open the pages behind it. If those pages list the competitor and leave you out, you have found the cause, not just the symptom.

Check what is said about you as well. OpenAI warns that search results and citations "can be incomplete, outdated, or incorrect," and Google says Gemini can hallucinate and present inaccurate information as fact. If an answer gets your pricing, your product or your market wrong, note the claim and the page it came from.

Hour four: tell a real gap from noise

Answers vary from run to run. OpenAI's API documentation says model outputs may differ from request to request, and the Gemini app lets you regenerate a response and switch between versions. One missing mention is a sample, not a finding.

A gap is real when it passes these tests:

  1. It repeats. Rerun the prompts where you were missing, in a fresh session. If you are still missing, keep it.
  2. It spreads across prompts. Missing from one of five prompts is weak evidence. Missing from four of five is a pattern.
  3. It is consistent by engine. Missing everywhere and missing on one engine are both real, but they have different causes. A gap only on the Google surfaces points at the pages Google ranks for that topic.
  4. The sources explain it. If the cited pages list competitors and not you, the gap has a cause you can act on.

Noise looks different: named in one run and not the next, a different brand in fifth place each time, or an AI Overview that appears one day and not the next.

The ten minute technical check

Before blaming your content, rule out access:

  • ChatGPT: OpenAI says sites that opt out of OAI-SearchBot "will not be shown in ChatGPT search answers," and recommends allowing it in robots.txt (OpenAI).
  • Perplexity: Perplexity recommends allowing PerplexityBot in robots.txt so your site can appear in its results (Perplexity).
  • Google: to appear as a supporting link in AI Overviews or AI Mode, a page must be indexed and eligible to be shown in Search with a snippet (Google Search Central). Search Console also has a Search generative AI control, where Include is the default (Google). Check nobody has switched it to Exclude.

If all three are in order, the answer to "why not us?" is in the sources you read in hour three, not in your robots file.

What the audit should produce

One page per topic, with:

  • how often you are named and how often your pages are cited, engine by engine
  • the brands named instead of you
  • the pages behind those answers, sorted by type, and whether you appear on each
  • anything the engines get wrong about you, with the page it came from
  • your Google ranking beside the answers

Resist the urge to boil it down to one number. Our guide to the AI visibility score explains what a single figure hides. And remember where this leaves you: with a diagnosis. The fix is getting onto the pages the engines already cite, a step an audit on its own does not cover.

The shortcut

If you would rather not spend the afternoon, run a free AI visibility audit with our checker. It drafts five buyer prompts from your site, you confirm them, and about two minutes later you get 20 answers from Google AI Overviews, ChatGPT, Gemini and Perplexity, with the pages each engine cited, the brands named instead of you, your Google position for the keyword behind the prompts and the cited pages you could be placed on, scored for difficulty. It is free, with no card and no email.

That covers hours one to three for one topic. The rerun in hour four is what daily tracking is for: alwaystracked checks 50 prompts and 10 keywords every day from $129 a month, with a 14 day free trial. If you want the gaps closed rather than reported, alwaysmentioned (from $1,395 a month) and alwayscited (from $3,395 a month) secure placements and link insertions in the pages the engines cite, with citation reporting on each one.

Questions

How long does an AI visibility audit take?
About an afternoon for one to three topics if you run it by hand: an hour to write the prompts, an hour to run them, an hour to read the sources and an hour to rerun the gaps and write up what you found. A free checker can do the first three steps for one topic in a few minutes.
What should an AI search audit include?
A fixed set of buyer prompts, the full answer from each engine, the brands named and in what order, the pages each answer cited, anything the engines get wrong about you, your Google ranking for the matching keyword, and a quick check that the AI crawlers are allowed to reach your site.
How do I tell a real gap from noise?
A real gap repeats. It shows up again when you rerun the prompt in a fresh session, it appears on most of the prompts in a topic rather than one, and the cited sources explain it, for example roundups that list competitors and leave you out.
Is an AI visibility audit worth paying for?
The diagnosis is the cheap part, and you can do it yourself in an afternoon. What costs money is the work that follows: getting onto the pages the engines cite. Pay for that work, and expect the audit to come with it.
Free scan

Check your AI visibility free

See which buying questions ChatGPT, Perplexity, Gemini and Google's AI answers name you for, and who they name instead. No call, no card.