What is an AI mention score?

Questions, not mentions. If ChatGPT and Perplexity both name you for "best roofer in Leeds", that's one question where you were named. Counting questions keeps one popular question from inflating the number.

The questions are the ones a buyer types without any business names in them: who's the best, who's cheapest, who handles this particular job, who's near me. A question that already contains your name tells you nothing about whether a stranger would hear of you.

Because one answer is noisy. The same question asked twice can name different businesses, and each assistant has its own list. A score across dozens of questions and four assistants evens out the luck in any single answer.

It also shows how spread out the answers are. In our Richardson, TX accountant study (September 7, 2026), 152 firms were named across 192 answers and 123 of them by only one assistant.

It depends on the city and the trade. In a big city with hundreds of firms, being named for a handful of questions can put you near the top. In a small town with a few rivals, you might aim for most of them.

Compare yourself with the businesses that get named most in your own report. And watch the trend: a score that moves from 3 to 9 over two months says your changes are being read.

Share of voice is your slice of every recommendation made across the run. If the four assistants made 400 recommendations and 20 were you, your share is 5%. The mention score counts questions; share of voice counts recommendations. We show both, because a business can be named often on a few questions or once each on many.

Yes. The score is built from the city you give us, so it works for a dentist in Dublin as well as one in Dallas. The questions are in English.