How to measure AI visibility each month

There is no Search Console for AI assistants. If you want to know whether ChatGPT, Perplexity, Gemini, Claude or Google AI Overviews name your business, you have to ask them, the same way, on a schedule. This is the method we use, and you can run a version of it yourself.

1. Write the questions before you look at the answers

Draft 20 to 50 questions a buyer would really ask. Split them into three groups:

• Branded: questions that name you ("What does [company] do?").
• Category: questions that name your service and market but not you ("Which agencies in Norway offer answer engine optimisation?").
• Decision: comparison and shortlist questions ("Best AEO agencies for Nordic B2B firms?").

Branded questions tell you whether assistants describe you accurately. Category and decision questions tell you whether new buyers ever meet you. Keep the two scores separate; blending them hides the gap that matters.

2. Record a dated baseline

Run every question before any work starts. For each answer, record three things: whether you are named, whether your site is cited as a source, and which other firms and sources appear. Note the assistant, model setting and date. A baseline taken after the first changes ship is not a baseline.

3. Re-run on a fixed cadence

Monthly is enough for most firms. Use the same questions and settings each time. Assistants are non-deterministic: the same question can return different sources on different days, so a single run is an anecdote and a trend across months is evidence.

4. Report what moved and what did not

• Named rate and cited rate for category and decision questions, against the baseline.
• Which third-party sources appear most often. Those are your corroboration targets.
• Errors assistants make about you, such as wrong services, prices or location.

Say plainly when a change is inside the noise. Citation share is a leading indicator; it can move before enquiries do, and it is not a revenue report.

5. Watch the measurement itself

Tools fail quietly. Different model settings can return sources in different formats, and a script that reads only one format will report zero citations when the answer was well sourced. Spot-check the raw answers whenever a number changes sharply.

Where this fits

This measurement is the backbone of an AEO and GEO engagement. If you would rather we ran it, the Free AI Audit starts with exactly this test across the five assistants and your buyers' questions, delivered within two working days.

Related reading

• AEO and GEO in Norway: a guide for Nordic B2B buyers
• Digital PR and AI citations belong in one plan
• Does schema help you appear in AI answers?
• How to evaluate an AEO agency: seven questions