Sampled measurement, confidence tiers, and what every number on your dashboard actually represents.
We do not measure search engine rankings. Instead, we sample answers from each tracked AI engine against your prompt set, then parse every answer for brand mentions, competitor mentions, position in answer, and cited sources. Every metric on your dashboard is an aggregation over those analysed answers.
Sample size matters. Every metric carries a confidence label so you know how much to trust it:
Built on a small number of analysed answers. Useful for spotting patterns, not for board reporting.
Enough analysed answers for trends to be meaningful; absolute numbers still move with new samples.
A deep sample behind the number. Figures and trends are stable enough to make decisions on.
Report directional reads as 'early signal'; wait for moderate+ before recommending material spend.
An analysed answer is one verbatim response captured from a tracked engine for a tracked prompt, parsed for entities and citations, and stored with timestamp, market, and language metadata. Answers that fail content extraction are excluded from sample counts.
We currently track ChatGPT (OpenAI), Perplexity, Google Gemini and Claude (Anthropic). Each engine is sampled independently; per-engine confidence is reported separately so you can see where the picture is clearer.
Default scan frequency is daily on paid plans and weekly on trial. You can trigger a re-scan manually from Settings → Brand at any time (subject to plan rate limits).
AI answers are non-deterministic. That's why we measure with sample-based confidence, not single-snapshot rankings.
AI answers are non-deterministic - the same prompt can yield different answers across samples. That is precisely why we measure with sample-based confidence rather than reporting a single "rank". Use directional readings to spot trends; wait for moderate or strong confidence before making material content or PR decisions.