What we measure
Whether AI names your business when customers ask the questions that come before a hire: “who’s a good roofer near Mason?”, not “what is your company?”. For each answer we record whether you’re named, where you sit in the list, and whether what it says is right.
The questions
Before work starts we agree a fixed set of buying questions with you: the jobs you want more of, the towns you serve, phrased the way customers phrase them. Then the set is frozen. If we change one, the report says so, because a changed question breaks the month-to-month comparison.
The engines
ChatGPT, Gemini, Perplexity and Google AI (AI Overviews and AI Mode). Every report lists which engines were checked and on which dates.
How the answers are collected
Every question goes to every engine in the same window each month, in a clean session with no account history. We save the full answer and a dated screenshot.
How it is scored
- Appearance rate: the share of questions where AI names you at least once.
- Share of voice: your mentions divided by every business mention across the answers.
- Position: where you sit when AI gives a list.
- Accuracy: whether the facts AI repeats about you (services, area, hours, reviews) are right.
What it can’t tell you
One month is a snapshot. Small moves are often noise, so we read the trend. A citation is not a recommendation, and a recommendation is not a booked job. We report each separately and never add them into one flattering number.
The free AI Visibility Score is different
The free score is a one-time check: five questions (two that don’t name you, three that do) across ChatGPT, Claude and Perplexity, scored out of 100. It’s a starting point. Client tracking uses your agreed question set across all four engines, monthly.