How to read an AI-visibility report without fooling yourself

The AI-visibility market is young, and young markets reward confident numbers over honest ones. A vendor can put '82% visibility' on a slide and most buyers will nod. And the stakes are real: EAB found 18% of students have removed a school from their list based on AI results alone. Before you trust a number, three questions separate a measurement from a decoration.

One: what is the denominator?

A percentage without a sample size is a number without a scale. '82%' out of eleven prompts is a coin flip dressed as a fact; the same figure out of two thousand prompts means something. Every honest visibility number should arrive with its n attached, and you are entitled to ask for it before anything else.

Two: which questions, and who chose them?

Visibility on your own brand name is easy and nearly meaningless — of course the engine knows you when asked about you. What matters is visibility on the category question a real buyer types: 'best pathway program for international students', not 'is Acme good'. If the prompt list is hidden, or curated to flatter, the number is theater.

Three: can you see the answer behind the number?

This is the one most tools fail. A trustworthy report lets you drill from any metric down to the AI's verbatim reply — the exact words, the engine, the timestamp — so you can check that the tool counted a real mention and not a coincidence of phrasing. If you cannot get from the score to the sentence, you are being asked to trust a black box.

None of this requires a statistics degree. Denominator, question set, verbatim evidence — ask for those three, and most of the market's confident numbers quietly get more honest.