How do you measure AI visibility?
You measure AI visibility by asking AI assistants the same set of non-branded category questions on a schedule and recording, for each dated answer, whether your brand is mentioned and whether your site is cited.
- Choose category questions a customer would ask that do not contain your brand name.
- Ask the same questions to more than one assistant: ChatGPT, Claude, Gemini, Perplexity and Google AI Overviews.
- Count brand mentions and citations of your own site separately; one does not imply the other.
- Calculate your share against the competitors in the same answers (Share of Voice).
- Repeat with the same questions and record the date. A question that returned no answer counts as "not measured", not as 0.
Exact metric definitions are in the SOV Tracker Methodology.
Why AI visibility matters
People increasingly use AI assistants alongside search, social networks, review sites and other discovery channels when researching products, services and organizations. AI answers can summarize only a small subset of available information, so brands benefit from measuring what those answers actually say instead of assuming that search rankings automatically translate into AI visibility.
You will also see the same idea called LLM visibility, AI search visibility or generative engine visibility. These terms describe the same measurement: how a brand appears in answers from AI assistants such as ChatGPT, Claude, Gemini, Perplexity and Google AI Overviews.
AI visibility is not a single score
Different measurements answer different questions. Combining them into one generic /100 number makes it harder to understand what changed. SOV Tracker therefore keeps the main answer metrics separate.
- AI Share of Voice: competitor-relative share of valid mentions on a compatible configured comparison scope.
- Discovery Visibility: found / answered on eligible non-branded discovery questions.
- Recommendation Rate: defined in the methodology but not measured yet; reports show it as not measured, never as 0.
- Owned-Site Citation Coverage: observed citations to the brand's own site on the eligible scope.
- Sentiment: bounded positive / neutral / negative classification.
- Prominence: where the first valid mention appears inside the returned answer; not a universal ranking.
- Brand Accuracy: whether eligible answer evidence represents business-confirmed brand facts accurately.
Where Share of Voice fits
SOV is useful when a real competitor universe exists. The canonical formula is target valid mentions divided by target plus configured competitor valid mentions on a compatible measurement scope. Prominence, sentiment, engine market share, question popularity and question volume do not modify that percentage.
What the Free Scan measures
The Free Scan is a limited ChatGPT check using five automatically generated non-branded questions. Its headline reading is Discovery Visibility. It does not provide continuous monitoring and it does not manufacture competitor-relative SOV when no valid competitor universe exists.
Paid measurement across five engines
Paid SOV Tracker plans cover ChatGPT, Claude, Gemini, Perplexity and Google AI Overviews. Priority questions receive a daily ChatGPT + Google AI Overviews pulse, while the full tracked set is measured weekly across all five engines. Canonical SOV comes from the synchronized full measurement window so incompatible cadences are not blended into one percentage.
How to improve the evidence
GEO work can include answer-first content, citable facts, consistent entity data, technical crawlability, authoritative third-party references and keeping important business facts current. These are optimization inputs; none guarantees a particular AI answer or citation. Measure before and after on compatible scopes.
For exact definitions and unavailable-state rules, read the SOV Tracker Methodology.