AI search visibility is easy to overstate. A screenshot proves that one answer appeared at one moment. It does not show how often the brand appears, which questions matter, whether the answer cites a useful source or whether anyone moves towards a commercial action.
Start with a stable question set
Choose questions that reflect real research and buying behaviour. Keep the set small enough to review consistently. Record the interface, location, account state and date because answers can vary.
Separate presence from citation
A brand can be mentioned without a link or source. Track at least three states: absent, mentioned and cited. Add a fourth state when the cited page is owned by the brand.
Record answer coverage
Note whether the answer covers the part of the question that matters commercially. A citation may be technically present while the brand is associated with the wrong product, market or claim.
Connect the visibility to behaviour
Use referral data where it exists, but do not stop there. Review branded search movement, direct visits to cited pages, assisted conversions and later sales conversations that mention AI research.
Report uncertainty
Interfaces change, personalisation varies and query volume is often unavailable. A useful report explains the method and the limit. It does not turn a small sample into a market share claim.
The goal is a repeatable view that supports content and evidence decisions. It is not a single visibility score designed to look certain.