Read your AI Visibility score and its confidence range
What the score is made of, what the 95% range means, and how to read the engine table and the two leaderboards.
The big number on the AI Visibility page is a 0 to 100 score for the week that just ended. Underneath it is a plain sentence about what that number means for you, from "AI does not know you are there yet" at the bottom to "AI already sends customers your way. Protect it." at the top.
What the score is made of
The formula is published on the page, at the bottom, and it does not change between the free scanner and your dashboard:
Score = 100 × (0.6 × visibility + 0.3 × citation + 0.1 × coverage)
- Visibility is the share of answers that named you or cited you. It carries most of the weight, because being named is the point.
- Citation is the share of answers that cited a page you own. Being the source is worth more than being mentioned.
- Coverage is the share of your questions that produced any answer at all. It is the smallest weight, because a question with no answer says more about the engine than about you.
One mention per answer counts once, and one domain per answer counts once, so an answer that names you five times does not inflate the score.
An answer that comes back empty, for example a query where Google shows no AI Overview, counts as no answer rather than an answer that left you out. It cannot lower your score.
The 95% range
Under the score you will see something like "95% range 23 to 33, from 100 sampled answers. More samples narrow it."
A weekly scan samples a few dozen answers, not a few thousand, and the engines are not deterministic: the same question asked twice can come back differently. The range is the honest width of what those samples can tell you. Two weeks are only really different if their ranges barely overlap.
Each question carries its own version of the same idea: the named chip shows the chance the next answer names you, with its range, for example "named 40% (12 to 74)". A wide range means it has only been asked a few times.
The six metrics beside the score
- Answers that named or cited you
- Answers that cited your site
- Questions that got an answer
- Share of voice vs tracked competitors, which stays blank until you track competitors
- Average position when named, your mean rank among the brands an answer lists
- Visits from AI assistants, 30 days, real people arriving on your hosted pages from an assistant
Engine by engine
The Engine by engine table breaks the same numbers out per engine. Engines disagree, often sharply: it is normal to score well in ChatGPT and near zero in Google AI Overviews. Read the low ones as the work still to do rather than as a mistake.
The two leaderboards
Who AI names instead of you is the list of businesses the answers recommended. Press track next to a name to start counting share of voice against it.
Where AI gets its answers is the list of sites the answers were built from, labelled as a competitor, a directory or portal, or a third party. Several of them have a How to get on fold-out with the steps and the link, because getting listed on a source the engines already trust is usually faster than outranking it.
Why the score moves
Week to week movement of a few points is usually sampling, not a change in the world. Look at the Score by week bars rather than at a single week, and treat a move as real when it is larger than the confidence range or when it holds for two weeks.
Movement that is real usually has a cause you can name: a page you published, a directory profile you fixed, a competitor who published something better, or an engine changing how it answers.