How to Build an AI Visibility Prompt Set for Your Business

Why a fixed set matters

Assistants are non-deterministic — ask the same question twice and you may get different names. That makes any single answer useless as a measurement, and a large fixed set essential. The stability comes from the aggregate, not the individual response.

Cover the stages a buyer moves through

Discovery ("who are the best X in Sydney"), comparison ("X vs Y for a small business"), decision ("how much should I pay for X"), and validation ("is X a reputable company"). Businesses often appear at one stage and not others, and the gap tells you what to fix.

Include your locations and services explicitly

"Best SEO agency in Chatswood" and "SEO agency for multi-location businesses" test very different things. If you serve multiple areas or sell multiple services, each needs coverage or you'll miss where you're invisible.

Add competitor-framed prompts

"Alternatives to [competitor]" and "[competitor] vs" reveal whether you appear in consideration sets you should be in. This is often the fastest-moving metric once entity work lands.

Record description, not just presence

Being named is necessary but not sufficient. How you're described matters — if an assistant consistently describes you as "a small agency" when you want to be seen as specialist, that's a positioning problem you can fix with content.

Run it the same way every month

Same prompts, same order, fresh session, no personalisation, logged with the date. Consistency is what makes the trend meaningful. It takes about an hour a month by hand, which is why we automate it as part of the AI visibility programme.

Act on the gaps, not the score

The number is only useful as a direction. What you act on is the pattern: which prompt types you lose, which competitors keep appearing, and what those competitors have that you don't. Usually it's third-party corroboration. See structuring pages so AI will quote them for the on-site half of the fix.