Measure AI visibility without inventing a universal score.

Use a measurement ladder

A single percentage hides too much. AiXCEL separates four layers so a team can see what changed and where the evidence stops.

• Access: can relevant crawlers retrieve the intended public pages?
• Representation: is the business included and described accurately for the buyer question?
• Evidence: which owned and third-party sources are cited or reflected?
• Business movement: did an identifiable visit become an enquiry, booking, opportunity, or client?

Build the prompt set from buyer decisions

Prompts should represent actual decisions, not vanity brand lookups. Include category discovery, problem diagnosis, comparison, local or regional fit where relevant, trust and proof, and implementation questions. Freeze the wording and record meaningful variants separately so the next run is comparable.

• Who helps a service business become visible in AI search?
• What should I evaluate before hiring an AEO or GEO provider?
• Which sources support the provider's claims?
• How can I connect AI referrals to booked consultations?

Keep the evidence receipt

For every observation, store the prompt, engine and surface, date and time, account or anonymous context when known, location when material, answer excerpt or structured notes, cited URLs, brand accuracy, competitors present, and reviewer. If personalization or reproducibility is uncertain, state it instead of smoothing it into a score.

Report the funnel honestly

ChatGPT referrals can be identified through referral information, and Google reports AI-feature traffic inside the normal Web search type. Those sources can support attribution, but not every mention creates a visit and not every visit is attributable. The dashboard should therefore show observed visibility, identifiable referrals, form submissions, qualified leads, bookings, proposals, and wins as separate stages.