AI visibility measurement should preserve the prompt, context, date and observed answer. One citation does not establish durable coverage or commercial impact.
Use a repeatable protocol
For a fictional product studio, select branded, comparison and project-planning questions. Distinguish web-assisted answers from responses without visible sources. Record actual linked URLs rather than counting a similar company name as a citation of your site.
| Observation | Record |
|---|---|
| Question | Exact wording and language |
| Context | Product, search mode and date |
| Presence | Mention, citation or absence |
| Source | URL actually linked in the answer |
| Follow-through | Observable visit or enquiry, when available |
Interpret variation carefully
Answers vary across contexts and systems. Repeat a stable protocol without claiming it represents every user's experience. Keep manual observations separate from available referral traffic. More mentions do not automatically mean more qualified enquiries.
Can a special tag guarantee inclusion?
No. For Google's AI search features, ordinary access and content-quality fundamentals remain relevant; special markup does not guarantee selection. Make information verifiable, sources accessible and organisational identity consistent. Then measure actual outcomes rather than presenting technical eligibility as proof of recommendation.
Keep branded and unbranded prompts separate
A system may find a company when explicitly named without recommending it for an unbranded buying question. Track those observations separately across languages. They answer different questions about discoverability and should not be combined into one impressive but ambiguous visibility score.
Sources and further reading
From this guide to your project
Growth ↗
A product finds its market when acquisition, activation and retention work together.
Tell us about your ideaExamples are educational and do not describe client results. Content prepared with AI assistance; see our editorial policy. Our method ↗