Make every visibility number explainable
Ask what was sampled, what completed and what evidence supports each reported value.
Start with scope and completion
A question set should reflect real buying needs without putting the brand in every question. Record the model route, date and whether answers use retrieved web material or model memory. A branded recognition question answers a different question from spontaneous discovery.
| Result | Interpretation |
|---|---|
| Completed, brand absent | A valid negative mention observation for this sample |
| Failed or missing | Unavailable evidence; exclude from a measured rate |
| Brand named with no rank | A mention, not an inferred recommendation position |
| Citation to another site | Source usage, not a citation to your website |
Keep recognition out of discovery scoring
Keep unprompted discovery scoring separate from answers to questions that already name the brand.
Discovery question: Which tools help a small team create reusable prompts from its product documents? Recognition question: What does [the named business and website] do? Save both answers if useful, but score spontaneous discovery only from the first. The second question supplied the brand and tests whether the model can describe that specific business.
Small samples need the answers beside them
A high percentage from a few answers has limited precision. Keep exact evidence accessible and retain dates. Compare only compatible questions, models, markets and methods; mark a changed panel instead of drawing a misleading continuous trend.
This methodology concerns saved API observations. It does not claim direct consumer-app monitoring, all-market coverage or a universal search position.
Common questions
Does a citation prove endorsement?
No. It establishes that an answer references a source. Read the surrounding text to determine whether it describes, criticizes, compares or recommends the business.
Sources and review
Reviewed