AI visibility audit: a step-by-step method
TL;DR
- An AI visibility audit measures whether your brand is cited by AI engines - ChatGPT, Perplexity, Gemini and AI Mode - and where you're losing to competitors.
- There's no Search Console for AI citations: you measure it on purpose, with a repeatable prompt baseline, share of voice and analysis of the cited sources.
- Because each engine cites different sources (only ~10% overlap), auditing just one gives a misleading picture - you have to cover all four.
- It's the R - Results tracking - step of my CITAR method: without measurement, GEO is faith; with it, it's management.
Everyone wants to "show up in ChatGPT". Few measure whether they do. An AI visibility audit answers the question that matters - are you cited, for which questions, and who beats you? - with method, not a screenshot taken on a lucky day. It's the work that turns GEO from faith into management.
What an AI visibility audit is
It's the process of measuring, repeatably, whether and how your brand is cited by generative AI engines. Instead of positions on a SERP, you look at citations inside answers: how often you're mentioned, for which questions, with which sources, and how you compare to competitors. GEO today is where SEO was in 2008 - everyone knows it matters, but measurement barely exists. The audit is what makes it possible.
Step 1: the prompt baseline
Start by fixing a set of real questions from your domain - the ones a client would ask - and testing them periodically in each engine, recording whether you're cited and where in the answer. The secret is consistency: the same questions, the same engines, the same cadence (monthly, say). Only then do you compare over time instead of reacting to one isolated result.
Step 2: share of voice and cited sources
Next, measure your share. For each question, who gets cited? If your competitors show up and you don't, you have an entity or citable-content problem. Also analyze which sources the engine uses - often specific pages (guides, comparisons, Reddit) you can work on. This is where the audit stops describing the problem and starts pointing to the fix.
Step 3: the Google side (GSC) and traffic (GA4)
Add quantitative data to the qualitative read. The Search Console generative AI report gives you impressions in AI Overviews and AI Mode; GA4, with a channel and a referral regex, gives you the sessions AI already sends. Join the three and you get an honest picture: visibility (prompts), impressions (GSC) and traffic (GA4).
| Layer | What it measures | Source |
|---|---|---|
| Prompt baseline | Whether you're cited, by engine and question | ChatGPT, Perplexity, Gemini, AI Mode |
| Share of voice | Your citation share vs competitors | Answer analysis |
| Cited sources | Which pages the engine reuses | Answers with citations |
| Impressions | Visibility in AIO and AI Mode | GSC (AI report) |
| AI traffic | Sessions and conversions from AI | GA4 (regex) |
How often to audit
Monthly for most businesses; fortnightly if you're in a competitive sector or running an active GEO plan. What matters is regularity - a one-off audit is a snapshot; the time series reveals whether you're gaining or losing ground.
How I use this
This audit is the R (Results tracking) step of my CITAR method. It closes the loop: measure, show where you lose, and feed the entity and citable-content work that makes you win. It's what I do in the GEO service.
An honest read
There's no single official AI-citation metric, and model outputs vary between sessions and users - which is why baseline discipline (same questions, same cadence) is what gives reliability. Nobody guarantees a citation; what the audit guarantees is that you stop guessing and start deciding on data.
Key data
| Engines to cover | ChatGPT, Perplexity, Gemini, AI Mode |
|---|---|
| Sources shared across engines | ~10% |
| Google-side signal | Generative AI report (GSC) |
| AI traffic | GA4 (referral regex) |
| Single official citation metric | None |
There's no single official AI-citation metric; combine a prompt baseline + GSC + GA4.
Sources: Google Search Central - Gen-AI performance reports · Adapt - AI search roundup 2026
Common mistakes and how to do it right
An AI audit is worth its method, not a screenshot. What makes it reliable:
| Avoid | Do |
|---|---|
| ✗ Asking once and drawing conclusions | ✓ Repeating the same set of prompts periodically |
| ✗ Changing the prompts every test | ✓ Fixing the questions to compare over time |
| ✗ Auditing ChatGPT only | ✓ Covering ChatGPT, Perplexity, Gemini and AI Mode (distinct sources) |
| ✗ Trusting perception alone | ✓ Cross-checking with impressions (GSC) and AI traffic (GA4) |
Tools I use and recommend
- ChatGPTAI assistant for research, analysis and quick audits.
- PerplexityAnswer engine with sources, useful for research and GEO.
- GeminiGoogle's AI assistant, grounded in Search.
- Google Search ConsoleOrganic performance, coverage and field Core Web Vitals.
- Google Analytics 4Behaviour and conversion measurement.
Sources
- Google Search Central - Gen-AI performance reports
- Adapt - AI Search Roundup (Jun-Jul 2026)
- Search Engine Land - zero-click searches 2026
Frequently asked questions
Do I need paid tools to audit AI visibility?
Not necessarily. A disciplined prompt baseline + the GSC AI report + GA4 already give you the essentials. Dedicated tools help you scale, but method comes first. GSC AI report →
Why isn't auditing ChatGPT enough?
Because each engine cites different sources - only about 10% overlap. Being in ChatGPT doesn't put you in Gemini or AI Overviews. AI Overviews vs AI Mode →
How often should I repeat the audit?
Monthly for most; fortnightly in competitive sectors or during an active GEO plan. Regularity is what makes the series valuable. GEO service →