How to Measure AI Visibility

Short answer: AI visibility is measured in four steps. Step one, build a real question set: not generic queries but 10-20 questions derived from your customer's buying journey (discovery: 'how to solve X', comparison: 'best X service', local: 'X in Izmir', brand: 'is company Y good'). Step two, ask them across ChatGPT, Gemini and Perplexity and record three numbers: mention rate (in how many questions you appeared), share of voice (your mentions relative to competitors'), average position (how far down the list you were). Step three, citation analysis: when the model mentions you, what source did it lean on — your site, a directory, a forum? Step four, monthly repetition: one measurement is a photo, the trend is the film. Scoring zero on the first run is normal for new brands; the real intelligence is seeing which sources feed your competitors — those sources are the target list for your content and presence plan.

AI visibility measures how often, in what order and in what tone your brand is mentioned in AI answers. The core metrics: mention rate, share of voice versus rivals, average position and citation analysis.

Build a real prompt set

Measurement starts not with generic questions but with questions derived from your customer's real buying journey. Split them into four intent layers: discovery, comparison, problem and branded.

The more realistic this set, the more meaningful the result.

Collect the right metrics

Turn the result into insight

A raw score means little on its own. The real value is seeing which competitors AI surfaces and which sources it draws from, then turning that into an action plan.

Repeating the measurement periodically is the only reliable way to see whether your improvements are working.

Our own first measurement, shared openly: on a five-question set our brand appeared only in the direct brand query and in none of the recommendation queries — the expected picture for a ten-month-old brand. That zero isn't bad news; it's the starting line: we now know which competitors get recommended and which sources the model cites, and we'll report the change monthly using exactly the metrics in this post.