What it means
Good buyer questions describe a real need, a type of buyer and constraints such as team size, budget, integrations or region, without naming the brand being measured. These unbranded questions show whether an engine chooses the brand on its own. Branded questions, which name it, test how accurately answers describe it.
A prompt set turns scattered spot checks into a measurement: the same questions, engines and settings run on a schedule, and results are compared only between runs that used them.
Why it matters
Every AI visibility number depends on the questions behind it. A set full of branded or easy questions flatters the result; a set that ignores real buying constraints misses the answers that decide a shortlist.
Example
A product analytics vendor builds 30 unbranded questions across discovery, comparison and pricing, such as 'Which product analytics tools work for a mobile app with no dedicated data team?'. It keeps five branded questions separate to check how answers describe its pricing and integrations.
Common misunderstandings
- Paraphrases of one question are not independent evidence of demand; they test how sensitive answers are to wording.
- A keyword list from an SEO tool is not a set of buyer questions; it lacks the context buyers give assistants.
- Results from different prompt sets cannot be compared as if they measured the same thing.