How to Measure AI Search Visibility

The five metrics

• Mention rate — how often your brand is named in relevant answers
• Citation rate — how often your domain is cited as a source
• Recommendation rate — how often you're endorsed as the solution
• Sentiment — how favorably you're characterized when mentioned
• Share of voice — your brand's slice of the total AI answer surface

Designing the prompt universe

Weight prompts by funnel stage — decision-stage prompts count more than informational ones. Include comparative and alternatives prompts explicitly; they're the highest-signal proxies for pipeline influence.

Size the universe to your category. Narrow B2B verticals hold a defensible baseline at 200–400 prompts; broad categories need 1,000+. A prompt universe that changes constantly produces noisy SOV numbers — lock it quarterly.

Cadence and alerting

Priority prompts should be sampled daily; long-tail weekly. Alert on any decision-stage prompt where a competitor overtakes you, and on any prompt where your sentiment flips negative.

Weekly summaries go to the operator team; monthly rollups go to executives. Quarterly reports tie SOV movement to specific program milestones — pages shipped, citations earned, entity properties confirmed. That's the loop that turns measurement into a management tool.

Common measurement mistakes

Treating classic SEO tools as sufficient. They don't sample generated answers; they can't produce AI visibility metrics.

Measuring citations without measuring mentions and recommendations. Recommendation is the highest-value outcome and it's not covered by any citation-only tool.

Running a rotating prompt universe. If the prompts change every month, the SOV number is measuring drift, not program performance. Lock the universe and version it explicitly when you change it.