Visibility Score
A visibility score is a single number from 0 to 100 that summarizes how often an AI assistant names a business when asked a fixed panel of the questions its customers ask. It is the reading this site takes of AI visibility: the state is the visibility, the score is its measurement.
How the number is built
It starts with the panel: twelve questions the business’s customers ask in some form every day, written the way people type them, with the abbreviations and typos left in. On this site each question goes to Claude, with live web search on, once per reading. For every answer three things are recorded: whether the business was named, whether it was named first, and who else was named beside it. Each answer is kept verbatim, with its date.
The score is then the share of the panel’s questions whose answer named the business, on a scale of 0 to 100: named in 3 of 12 is 25, so a business absent from every answer sits at the bottom and one named in every answer sits at the top. Being named first is reported beside the score rather than weighted into it, so anyone holding the answers can recompute the number. The methodology page walks through each stage.
Why the questions stay the same
The panel is fixed once it is written, and every reading is taken against it: the baseline, again once the fixes are live, and every month with monitoring. A question set that keeps changing would make every reading incomparable with the last, and a score whose movement cannot be trusted is not a measurement. Comparing each later reading with the baseline is the whole point of having one.
One question people ask
Why does the score move when the website has not changed?
Because an assistant’s answers vary from run to run, and from month to month as it re-reads the web. That is why no single answer is read as a verdict and every reading is dated: one question moving is noise; a sustained shift across the whole panel is a change in how the assistant sees the business.
Related terms
Sources
The panel, the assistant, the reading and the score are described on this site’s methodology page. No third-party figures are used on this page.