An AI visibility audit measures whether AI assistants recommend your brand when buyers ask. You run it by asking ChatGPT, Perplexity, Gemini and Claude your real buyer questions, recording whether you’re named and which page (if any) gets cited, then diagnosing why: can the engines read your site, do your facts exist and agree, and are you cited on sources they trust? It’s the baseline every GEO program is measured against — and you can run a first pass yourself in an afternoon.
What an AI visibility audit really measures
Traditional SEO audits ask “where do we rank?” An AI visibility audit asks a blunter question: when a customer asks an assistant for a recommendation, are we in the answer? A Google results page gives ten brands a shot. An AI answer usually names one to three. So this audit isn’t a softer version of an SEO audit — it’s measuring a different, harsher game where second place often means invisible.
The good news: the first pass is something you can do yourself, today, with nothing but the assistants themselves and a spreadsheet.
Step 1 — Ask the engines your real buyer questions
Don’t ask “is [my brand] good?” — that’s not how customers search. Ask the way a buyer who has never heard of you would. For each of your top services or products, run questions like:
- “What’s the best [your category] for [your customer type]?”
- “Who should I hire for [your service] in [your city]?”
- “[Top competitor] vs alternatives — what else should I consider?”
- “Recommend a [category] that [your key differentiator].”
Run each across the four engines that matter for commercial intent: ChatGPT (with search on), Perplexity, Google Gemini / AI Overviews, and Claude. Each behaves differently, so test them separately.
Step 2 — Record what comes back, precisely
For every question and every engine, log four things in a spreadsheet:
- Were you named? Yes / no. This is the headline number.
- Who was named instead? Your real AI competitors are whoever shows up here — often not who you think.
- What was said about you (if anything) — and is it accurate? Outdated or wrong facts are their own problem.
- Which page got cited? Perplexity and ChatGPT Search show sources. Note the exact URL — it tells you what’s already working.
Run each question two or three times; answers vary. You’re looking for patterns, not a single verdict.
Step 3 — Diagnose why you’re missing
If you’re not getting named, it’s almost always one of three failures, in this order:
- The engines can’t read you. Most AI crawlers don’t run JavaScript. If your site builds its content in the browser, they see an empty shell. Check by viewing your page’s raw HTML (right-click → View Source, or fetch it without scripts). If your text isn’t there, nothing else matters until it is. We wrote the full diagnosis here.
- Your facts don’t exist or don’t agree. If your site never plainly states who you serve, what you do, where, and what it costs — or if your site, directories and profiles say three different things — the model hedges and leaves you out. Where AI gets its facts covers this.
- You’re not cited where the engine looks. Each engine trusts certain sources per topic. If your category’s trusted directories, comparison pages and publications never mention you, you’re not in the candidate pool.
Step 4 — Fix the weakest link, then re-test
Resist the urge to fix everything at once. Work top-down: crawlability first (it gates everything), then factual clarity and consistency, then citations and authority. After each change, re-run the same questions and watch the numbers. The audit isn’t a one-time document — it’s a baseline you re-measure. That trend line, not any single screenshot, is how you know GEO is working.
When to bring in help
The self-run audit tells you whether you have a problem and roughly where. Fixing crawlability on a JavaScript site, building cross-web consistency, and earning category citations is ongoing work — that’s what a GEO agency is for. But you should never hire one blind. Run the baseline first; it costs an afternoon and makes every later dollar accountable.
Questions people ask
An AI visibility audit measures whether AI assistants recommend your brand when potential customers ask for one. You ask ChatGPT, Perplexity, Gemini and Claude your real buyer questions, record whether you’re named and which page gets cited, then diagnose why you appear or don’t. It’s the baseline a generative engine optimization program is measured against.
Yes. A solid first pass takes an afternoon: list your buyers’ real questions, ask them across the four major engines, and log whether you’re named, who is named instead, what’s said about you, and which page is cited. The deeper technical diagnosis — crawlability, fact consistency, citations — is where most brands eventually bring in help.
Almost always one of three reasons, in order: AI crawlers can’t read your site because it renders in JavaScript; your facts aren’t stated plainly or contradict each other across the web; or you’re not cited on the sources that engine trusts for your category. Fix crawlability first — it gates everything else.
Want this done for you?
We run this audit — across every major engine, with the technical diagnosis and the fixes — as the first step for every brand we take on. Get yours free.
Get a free AI Visibility Audit →