Should AI visibility audits separate clean-session and personalized results?

Two runs of the same prompt can diverge because one is anonymous while the other inherits account history, location, or a selected search/grounding mode. If a dashboard blends those observations, the average looks precise but is not reproducible.

A practical split:

1. Controlled baseline — fixed prompt version, market/language, engine and mode; a fresh session; timestamp; and the cited URLs.

2. Personalized diagnostic — explicitly tagged as a returning session or profile, with only scenario-level interpretation.

3. Coverage state — distinguish “not sampled” from “sampled, no mention” so missing runs do not silently become zeros.

For reporting, I’d compare baseline cohorts only with like-for-like runs. Personalized results can reveal edge cases, but they should not be generalized to the wider market. A small team could spend most of its sampling budget on the controlled baseline and reserve a few runs for personalized scenarios.

I’m building Corank (https://corank.ai), an AI-search visibility platform, and this is one of the measurement questions we keep coming back to. How would you represent session context without overwhelming a non-technical user?