How we measure
The AI Visibility Score is counted from real answers. Here is exactly how.
The questions
We read your website and write 6 questions a new customer would ask without knowing you: 4 asking for recommendations and 2 starting from a problem, with your city if you serve customers locally. Then 2 questions about you by name. Your questions stay the same for repeat checks.
The engines
Every question goes to ChatGPT, Gemini and Perplexity through their official interfaces with live web search switched on: 24 answers in total. We use the current standard models of each provider.
Reading the answers
We check every answer for your name, its common spellings, your website address and, for practices, the lead professional. Another model lists the businesses each answer recommends, in order. Answers that fail are left out of every calculation and shown as “no answer”.
The score (0 to 100)
Recommended (50): share of buyer answers that name you. Position (15): how high you appear when named, first place counts fully. Your site as a source (15): share of buyer answers citing your website. Accurate picture (10): whether the engines know you and state your facts correctly. Technical access (10): AI crawlers allowed, page indexable, business data readable. If a part can't be measured (for example because your website blocks automated visitors), it is left out and the rest is scaled to 100. If more than a third of answers fail, we show no score.
Limits
AI answers change between requests and over time. With identical questions we measured differences of about 3 points. A check is a snapshot, not a guarantee. We do not estimate lost revenue, because we can't measure it.