How we measure.
One screenshot proves nothing. AI answers vary by run, by time, by location and by model version. The measurement below is the only referee for the monthly settlement.
Sampling
20 or more fresh sessions per buyer question, per engine, per month. Each session starts clean. No prior context. No cached account state. Timestamps and engine versions are recorded on every run.
Locale pinning
Sessions are pinned to the agreed locale for the campaign. US English on US infrastructure by default. Regional variants on request. Locale drift invalidates a run.
Engines
OpenAI ChatGPT, Google AI Overviews, Anthropic Claude, and Perplexity Sonar. Each engine is settled independently. A miss on one is not a miss on another.
Answer Presence
We report Answer Presence as a rate. Named or not named. Position in the shortlist when named. Frequency across the sampled runs. Baseline is locked at day zero and is protected. Questions you already win cannot be sold to you as an outcome.
The run log
Every run is written to the log with prompt, engine, engine version, locale, timestamp, raw answer, and the extracted Answer Presence value. Logs ship with each monthly settlement and are re-runnable by you.
Standards
Truthful, bylined, disclosed. No fake reviews. No bots. No bought engagement. No guaranteed rankings, because anyone who promises those is guessing. What we guarantee is the measurement and the settlement rule.