What an AI visibility score is — and what it is not
A score is a summary of a sample. Here is exactly what ours samples, what it cannot see, and why we show a dash instead of a zero when we cannot measure.
Any single number describing something as large as "how AI sees your business" is a compression, and compressions lose things. That is fine, as long as you know what was lost. This is what ours keeps and what it throws away.
What the score is
It summarises a sample: a set of commercial questions, asked of a set of AI engines, at a point in time, with the answers recorded.
For a free check that is three questions across four engines. Each answer is examined for three separate things:
- Mention — were you named at all?
- Citation — was your site listed as a source?
- Recommendation — were you actually put forward as an answer to the buying question?
Those three are recorded separately because they move independently. Collapsing them would hide the case that matters most: cited as a source, recommended to nobody.
What the score is not
It is not a market share. No engine publishes how many people asked about your trade, so nobody — including us — can tell you what proportion of real conversations you appeared in. Any supplier quoting you an AI market share has modelled it, and should say so.
It is not a forecast. These systems change their answers without notice. A score is a reading, not a prediction.
It is not comparable across a methodology change. Our free check moved from five questions to three in August 2026. Scores from before and after that change are not comparable, and we treat them as different measurements rather than quietly plotting them on one line.
It is not a guarantee. We can show you what four engines said and which sources they trusted. Nobody can promise you a place in tomorrow's answer.
Why an unmeasurable check shows a dash
If a run cannot produce a usable reading — an engine refuses, a question comes back empty, retrieval fails — the honest output is "we could not measure this", not zero.
Zero has a meaning: we asked, and you were nowhere. A dash means we did not get a usable answer. Displaying the second as the first invents a bad result that we did not observe, and the customer has no way to tell the difference. So an unscoreable audit renders an em dash, and the run is labelled as unscoreable.
This costs us something. A dash is less satisfying than a number, and it makes the product look less certain than competitors who always print a figure. That trade is deliberate.
Why the questions matter more than the score
The score summarises; the questions determine what was summarised. A check that asks "who is the best plumber in Leeds" measures something quite different from one asking "emergency plumber near me open now". Both are legitimate; they are not interchangeable.
That is why the questions are visible before a check runs, and editable. A number derived from questions you never saw is not a measurement you can act on.
How to use it properly
- Compare against yourself over time, on the same questions, not against another business's number.
- Read the evidence. The answers and citations behind the score are the actionable part; the number is the headline.
- Watch the three components separately. Rising mentions with flat recommendations means you are becoming known without becoming the answer.
- Expect movement you did not cause. Engines update. Some variance is theirs, not yours.
Every number in a check links back to the answer it came from. If you cannot get from a figure to the text that produced it, be suspicious of the figure.