Start with the commercial question
Define the clinic, priority market, language, treatment lines, patient segments, and competitor set. The query framework should represent discovery, procedure, trust, aftercare, safety, value, and decision intent where relevant.
Freeze before observing
Changing prompts after seeing outputs destroys comparability. Freeze the wording and category map before the run, then execute multiple independent observations for every query.
Preserve evidence and review ambiguity
Retain raw responses within provider terms. Match known entities through verified names and domains. Route unknown clinics, unclear recommendation language, and uncertain source associations to manual review.
Disclose the limits beside the result
State provider, date, language, market, prompt count, runs, successful observations, formulas, and validation checks. Explain that the sample cannot prove clinical quality, causal revenue, or a permanent AI position.
Source and methodology note
This article explains the Medical AI Growth V1 benchmark definitions. It makes no scientific or clinical outcome claim.
Full methodology