How to measure AI visibility without inventing a comforting score

No single number captures visibility across every AI experience. Models, location, context and time all affect outputs. Use a scorecard, not a magic score.

Layer one: a fixed query sample

Build a set across discovery, comparison, specifications, objections and brand questions. Hold language, market, date and tool as stable as practical. Record brand mentions, links, citation context and answer quality.

Layer two: referrals

Monitor known referrers in GA4 and preserve UTM parameters where a channel allows them. Evaluate engagement, lead quality and orders—not session volume alone.

Layer three: Google Search

Reporting can change. Check the current official AI features documentation and your own Search Console interface before assuming a breakdown exists for your account and market.

Layer four: commercial outcomes

Use a consistent event model for service views, audit starts, submissions and purchases. Avoid attributing a lift to AI simply because it followed a technical change.

A minimum scorecard

• Brand presence across the fixed sample.
• Answers with a citation or link.
• Share of voice against defined competitors.
• Known referral sessions and quality.
• Attributable leads, orders and revenue.
• Qualitative accuracy of product descriptions.

Publish the limits

State sample size, measurement window, language, region and login state. Good measurement makes uncertainty visible so teams do not turn noise into strategy.