How do you measure AI visibility reliably?

“We agree a stable question set, surface, region, account context, date, and recording method. Each run captures mentions, citations, cited domains, factual accuracy, position or prominence where meaningful, and changes. Results are paired with search, referral, and conversion data. A prompt sample measures that sample, not every possible answer.”