How to build an AI visibility prompt set
Every number in an AI visibility programme inherits its prompt set, so a bad set produces confident, useless reporting. The common failure is choosing prompts by search volume, which pulls in high-traffic questions your brand could never credibly answer and then ranks them as your biggest opportunities.
What makes a good tracked prompt?
One where your brand appearing in the answer would genuinely help the person asking. If you would be an odd answer, tracking it produces a number that cannot be improved and should not be.
This sounds obvious and is routinely violated, because keyword tools reward volume. A privacy analytics company tracking a high-volume question about keyword research tools will score zero on it forever, and any ranking that weights volume will surface it as a top opportunity.
The five kinds of prompt worth tracking
Best tool for a job, in the phrasing a buyer would use. The highest-value and most contested.
Closest to purchase intent, and the easiest to win because the question already implies dissatisfaction.
Best option for a specific situation, budget or requirement. Less contested and often where a smaller brand can win outright.
Is your product any good, is it compliant, what does it cost. This is where you find out whether assistants describe you correctly.
The question a buyer asks before they know your category exists. Hardest to win and the earliest point at which you can be present.
How many prompts do I need?
Enough to cover the sub-questions in your category rather than a headline term or two. Coverage of the fan-out matters more than raw count.
Adding near-duplicate phrasings of the same question inflates the count without adding information, and costs real money on every scan. Breadth across distinct buying questions beats depth on one.
Common questions
Yes, but not as a visibility metric. Branded prompts tell you whether assistants describe you accurately, which is a different and often more urgent question.
Rarely. Changing prompts resets comparability, so treat additions as deliberate and keep the core stable.
Covers how AI visibility is calculated and where the numbers mislead.