Why it matters
Your customers don't type your name into an assistant. They type a trade and a town, and the answer is a short list that changes from one asking to the next. A single screenshot of one answer proves very little. A fixed list of questions, asked the same way again and again, builds a record you can compare from one month to the next.
That's why the wording stays frozen. If you reword a question, it's a new question with a new baseline. The rate you get from those tracked questions is what an AI visibility score boils down to one number. Prompt tracking is the questions and the log. The score is the summary.
Example
A plumber in a mid-size town writes questions around four things a buyer needs, an emergency, a replacement, a specific repair, and a price comparison. Each one gets a few phrasings, with the town in it, written the way a customer who hasn't heard of the business would type them. The plumber asks each question on each assistant on a set day every week and logs the date, the assistant, whether the business was named, who was named instead, and whether the site was cited. After a few months there's a trend line instead of a hunch.
