The score is not an expert's opinion: it is a reproducible measurement of your presence in AI answers. Here is exactly what we ask, how we weight it, and what the measurement does not say.
They are written from the real content of your home page — your services, your area, your market's vocabulary — then composed along fixed shapes: « best… », « comparison… », « where to find… ». The model does not write the questions, it only supplies vocabulary; the composition is done by the program, which is what makes two scans of the same site comparable. You can correct them: a query you add weighs as much as ours.
Each question is asked with no conversation history, no personalisation, in the language and country of the chosen market. One question, one clean session.
Perplexity, Gemini and Mistral are not queried in V1. We do not display them as measured until they are.
Three signals, measured across the 10 queries, brought back to 100.
Share of queries where your brand is named in the answer.
Where your mention falls in the text. Opening the answer is not the same as closing it. This signal only counts on answers where you appear, otherwise a single well-placed mention would be worth as much as twenty.
Share of queries where one of your URLs is cited as a source by the engine.
A scan is a photograph at a given moment. AI answers change from one week to the next, sometimes with nothing having moved on your side. That is why comparing two runs is done query by query, on frozen questions, and not on the overall score alone.
A placement is recommended on verifiable facts: how many times the AI already uses that domain as a source on your queries, and whether your competitors are on it. We display no forecast score gain — nobody controls what a model decides to cite. The real effect is read at the next run.
✓ The measurement: the same queries, the same engines, the same settings at every scan.
✓ The evidence: each engine's raw answer, dated, readable query by query.
✓ The legibility: no figure displayed that you cannot find again in the report.
— Model decisions. Nobody controls what an AI chooses to cite.
— A target score. We do not sell a promised number.
— Answer stability. Two scans a fortnight apart can differ with nothing having changed on your side.
We answer directly. Twenty minutes with our GEO expert, free, to read your data together and understand where your score comes from.
Book a call →