Platform · Intelligence
What AI engines answer, question by question.
We ask your buying questions to ChatGPT, Claude, Gemini, Perplexity and Google AI Overviews, several times, with frozen models. Every answer is kept and classified: your brand is named, cited only as a source, or absent.
What we measure
Three possible states for your brand in every answer, plus the competitors named.
The brand is in the text
The AI cites you by name in its answer. That is what the buyer reads.
Your site is used, without your name
Your page is cited as a source, but the answer does not name you. A gain still to convert.
Neither named nor a source
The answer is built without you. Often with a competitor.
Who appears in your place
The brands named in the same answers, recorded at every measurement.
Experts + proprietary technology
A measurement built to be rerun.
AI engines never answer exactly the same way twice. The protocol is designed so that two measurements are comparable.
- Choose the questionsWith you, based on your buying journeys and your competitors.
Experts + you - QueryEvery question, on every engine, 5 times, frozen models.
Measurement engine - KeepEvery answer and its sources, timestamped, delivered raw.
Measurement engine - ClassifyNamed, source only or absent; competitors named; tone of the answer.
Measurement engine - CompareSame protocol at every measurement. A change below ±15% is treated as noise.
Decision engine
The proof
What the measurement showed, in a real case.
US B2B SaaS vendor (anonymized) · August 9, 2026 → September 22, 2026. Same protocol at both measurements.
answers out of 1,000 named the brand at the baseline measurement
110 source only, 762 absent
answers naming the brand, at the next measurement
out of 1,000 answers at each measurement
on Perplexity, the engine that moved most
out of 250 answers
on ChatGPT, which did not move
we publish that too
Go further