METHOD NOTE 01Build a baseline you can compare.
Freeze the buyer questions, language and market first. Keep API model answers, search-enabled API answers, actual product interfaces and search AI summaries in separate groups. Preserve the original answer, sources and timestamp for each successful observation.
Always show the denominator. If 15 of 20 planned observations are valid and 3 mention the brand, coverage is 75% and the valid-answer mention rate is 20%. The five failures are missing observations, not five negative answers.
METHOD NOTE 02Make important company facts verifiable.
Separate company assertions, third-party documentation and independently verified facts. Product names, technical capability, certification and capacity statements should point to a source and a precise location in that source.
Keep conflicting evidence and dates. Mark an uncertain claim as unverified or disputed. Publishing the same claim in more places does not make it more reliable.
METHOD NOTE 03Measure inquiries separately from visibility.
A brand mention is an observation. A referral visit is a measured session. A qualified inquiry is a CRM decision supported by contact and demand information. These are different outcomes.
Retain first-touch, last non-direct touch and customer-reported sources separately. Unknown is a valid value. Do not convert a mention-rate increase into an assumed number of orders.