Consumer Showdowns use actual answers from named consumer AI interfaces under documented conditions, covering products, services and public-interest issues. API Lab studies use model endpoints and are separate. We do not combine those results.
For each Showdown we freeze the buyer scenario and prompt, document the interface, conditions and collection time, preserve the response and distinguish ranked recommendations from incidental mentions.
We verify material claims against appropriate primary sources where possible, report uncertainty and limitations, and treat recommendation overlap as different from accuracy or buyer fit. Research records preserve the prompt, conditions, timestamps, citations and results so studies can be repeated after days, weeks, months or years and compared historically. Publication requires human editorial approval.