How to measure your AI visibility?

How do you measure your AI visibility?

AI visibility is measured by crossing two things: real query tests on generative engines (are you cited, and where?) and a technical analysis of the site (do you meet the conditions to be cited?).

One without the other is not enough. Tests alone record an absence without explaining it; technical analysis alone predicts without verifying. The useful measure is the one that links every observed absence to a fixable cause.

Which queries should you test?

The commercial queries your customers actually ask an AI: recommendations ("which solution for…"), comparisons ("X or Y"), and expertise questions tied to your industry.

Three to five well-chosen queries are enough to photograph your position: who is cited, who is not, and which sources hold the two or three spots available in each answer. The result is often blunt: in the Novotel audit, the brand appeared only once across three key queries, and its sister brand Sofitel was recommended in its place.

What is a good AI visibility score?

On the Repliq grid, scored out of 100 across six weighted dimensions: below 50, structural barriers block citation; between 50 and 70, the foundation exists but blockers remain; above 70, the site is well armed and the work shifts to fine-tuning.

• Novotel: 52/100 "Fair", with solid brand authority (70) but weak citability (45).
• Figma: 73/100 "Good", with near-maximal authority (94) but AI crawlers restricted by an overly narrow robots.txt.
• In both cases the overall score hides internal gaps: the average matters less than the dimension-by-dimension diagnosis.

Why isn't a citation tracker enough?

A tracker counts your absences; it does not say why you are absent, nor what to fix.

AI visibility tracking platforms query models and count brand mentions: useful for following a trend, insufficient for acting. The audit reads your site the way an AI crawler would (access, structure, markup, authority) and turns every gap into a concrete action, with its effort and impact.

Frequently asked questions

How often should AI visibility be measured?

A complete audit takes the starting photo; after that, a monthly or quarterly follow-up is enough to measure progress and catch regressions, while fixes get recrawled by the bots.

AI answers vary: is the measure reliable?

Answers vary from session to session, but structural positions barely move: a well-armed brand comes back regularly, a technically blocked brand never appears. That structure is what the audit measures, not a single answer.

Can you measure a competitor's AI visibility?

Yes, the same query tests reveal who holds the citation spots in your industry. It's often the most persuasive part of a report: seeing a competitor cited where you are absent.