Tracked questions: the questions behind your mention rate
A tracked question is a question customers ask AI assistants, tracked so you can see whether your business gets mentioned in the answers. It’s the atomic unit of AI visibility measurement: every number in our tracking is built by asking these questions and reading what came back. This page explains the unit properly, because once it clicks, the rest of the measurement stops being mysterious.
A question, asked the way customers ask it
A tracked question is written in customer language, not marketing language: “best hvac company near me,” “who fixes water heaters in [city],” “dentist that takes my insurance.” It gets asked on a recurring basis across the AI assistants customers use, and each answer is recorded and read for one thing first: did it name your business.
The phrasing discipline matters more than it looks. Owners write “HVAC services [city]” because that’s how they’ve been taught keywords work. Customers type “my AC died who should I call.” Those two get different answers, and only one of them is the question your next customer is actually asking.
Core prompts and custom prompts
Two kinds, one difference: who chose them.
Core tracked questions are the questions we already know customers ask in your industry: the cost questions, the emergency questions, the near-me questions that recur in every market. They exist so tracking starts useful on day one instead of after a research project.
Custom prompts are questions specific to your business that you pick yourself: the service only you offer, the neighborhood you want, the question your phone calls always start with. Choosing them well is a skill worth ten minutes of thought, because a custom prompt nobody asks produces a number nobody needs.
Why questions and not keywords
Keywords were the unit when the output was a ranked list: a keyword had a results page, and you had a position on it. An AI answer has no position to hold. It’s a written response to a full question, and the same business can be named for “emergency plumber near me” and invisible for “best plumber near me” because those questions pull different sources and different reasoning. Measuring by question keeps what you’re told honest: you learn where you win and where you don’t, question by question, instead of averaging it into a ranking-style number that answers don’t actually have.
From prompts to your mention rate
Each tracked question gets asked, each answer gets recorded, and the yes-or-no of “were you named” rolls up into your mention rate: the share of tracked questions where AI named you. The companion numbers (citation rate and share of voice) come from reading the same answers for sources and competitor names. One prompt’s swing means little, because answers move week to week for their own reasons; the rate across the set over time is the number with a spine.
What a tracked question is not
Not a promise, not a submission, not a registration. Tracking a prompt doesn’t make AI mention you, the same way a thermometer doesn’t warm the room. Our opinion, briefly defended: that’s precisely what makes the unit trustworthy. A measurement you could buy your way into would be worthless as a measurement, and the entire value of watching tracked questions is that the answers are earned or absent, never purchased.