How the AI Visibility Score is calculated

It is one number, 0–100, for how visible a brand is in the answers AI assistants give to a set of buyer questions. It is presence multiplied by prominence: how often you are named across the period's answers, weighted by how visible you are in the answers that do name you. A brand named in every answer but always as an afterthought, and a brand named half the time but always first, can land in the same place — which is the point of combining the two rather than reporting either alone.

For every stored answer, one judge model scores five dimensions — coverage, placement, prominence, frequency and framing, in that order of weight — and our server blends them into that answer's visibility figure, applying fixed caps so an over-generous breakdown cannot inflate it. The headline score is those per-answer figures totalled across the period and divided once by the number of answers we scored. An answer that never names you contributes zero to that total rather than being dropped from it, which is exactly how presence stays inside the number instead of being reported separately. The model scores the components; the server does the arithmetic. No model ever hands us a headline number that we keep.

There is no universal threshold, and any tool that gives you one is inventing it. The score depends entirely on the questions you track: a narrow branded question will score high for almost anybody, and a wide category question is hard for a market leader. Two published measurements for scale: over seven days of one shopping question, Amazon scored 79, and over seven days of one ecommerce-platform question, Shopify scored 85 — both were named in all 21 answers. Read your own score as a baseline to move, and compare it with the competitor scores on the same board, which use the same denominator.

Because presence is only half of it. If you are named in 100% of answers but always last in a list of five, your prominence is low and the score follows the product of the two, not the presence figure. The five-dimension breakdown on the dashboard shows which half is costing you: coverage and frequency move with presence, placement and prominence move with where you land inside the answer.

Yes, and that is deliberate. Every competitor named in an answer is scored on the same five dimensions, blended by the same server-side arithmetic, over the same denominator. It is the only way a leaderboard means anything — a tool that scores you carefully and your rivals roughly is comparing two different measurements.

Yes. The score averages every tracked topic equally, so adding broader questions to a set of narrow ones will usually lower it on the day you add them, even though nothing about your brand changed. That is honest rather than convenient, so the dashboard marks the run where the topic set changed instead of hiding the step.