Evidence, not a score
Evidence, not a score is the principle that every number in an AI-visibility report must open back to the raw material it came from — the verbatim answer, its cited sources, and the run metadata — instead of a sealed, black-box composite the client has to trust.
A single "visibility %" merges signals that fail for different reasons and cannot be checked when a client asks what changed. Evidence-first reporting keeps the booleans separate and makes every derived rate, split, or trend traceable to a stored answer with a stated sampling window.
This principle also sets a hard boundary: the method proves what the engines said and what moved after a fix — it does not claim to make an engine say something, and it does not fact-check the model’s prose.
Related terms
Start with the free audit
Send us one brand. We’ll run it through the engines and send back the answers, citations, and sources — so the first report your client sees is already backed by evidence.
Free audit: ChatGPT, Perplexity and Gemini, run by hand. Paid work covers all six engines.