Measuring AI visibility

AI answers are not static: the exact same prompt can yield a different answer today than tomorrow. That makes tracking trickier than a standard Google rank check, but not impossible. You are tracking a trend, not an exact rank.

Below is the method I use for clients. You can adopt it yourself with nothing more than a spreadsheet and an hour a month.

Write down twenty to twenty-five questions just as a customer would type them, not how you would phrase them yourself. Mix three types: choice prompts ('which SEO specialist in Utrecht helps SMEs?'), pricing prompts ('how much does an SEO specialist cost per month?') and informational prompts ('SEO specialist or agency, which is better?').

Also include questions where your business does not currently appear, but where you want to be visible. These are your growth prompts; this is where the needle should move over the coming months.

One: are you mentioned, yes or no. Two: how are you described — is the description accurate and does it include facts like pricing or service area? Three: which other providers are mentioned and in what order.

That third point is often overlooked, yet it is the most actionable. If you keep seeing the same three competitors, you can inspect their pages and citations to see what type of source material is being picked up.

Keep it simple: mention rate per assistant, calculated as the number of prompts where you are mentioned divided by the total number of prompts asked. Alongside that, keep a quality score from one to three for how accurate the description is.

Always test the exact same way: the same questions, no previous chat history, and no logged-in session where the assistant already knows you. Otherwise, you are measuring your own past interactions rather than what a new customer would see.

If you are not mentioned in choice prompts but do appear in pricing prompts, your facts are clear but your authority is weak: focus on reviews and external citations. If you are mentioned with an inaccurate description, your own pages are either unclear or inconsistent with your external profiles.

If you see zero mentions across the board, do not start with AI optimisations—start with indexation: are you fully indexed in both Google and Bing? Nine times out of ten, that is where the problem lies.

Once a month is plenty; measuring more frequently mostly creates noise because models fluctuate daily. In practice, updates to facts and page structure usually feed through within a few weeks. New external citations take longer to show an effect, as they need to be indexed first.

Don't expect a straight line up. A healthy trend fluctuates but climbs over a quarter. One disappointing month is not a signal; two in a row is.

In my monthly report, you'll find the mention rate per assistant, the verbatim answers to five fixed key questions, the competitors mentioned, and one clear action for the following month. No dashboard packed with graphs, just a single page you can read in five minutes.

Those verbatim answers matter more than the percentage: they let you see firsthand exactly how a potential client is introduced to your business.