Written 14 October 2026.
There’s no single AI shortlist to rank in. Say on the page who your product is for, in the words that customer would type, and track your visibility one customer prompt at a time.
What I measured
On 9 October 2026 I asked ChatGPT and Gemini for the best accounting software in the UK three ways: with no persona, as a self-employed plumber, and as the finance director of a UK manufacturer with 250 staff. Each prompt ran twice per engine, in temporary chats.
Three of the four finance-director answers recommended none of the products the no-persona answers named. The plumber got the same three brands as the generic question, but ChatGPT changed its top pick from Xero to FreeAgent. Study and CSV.
What to write
[Product] is accounting software for self-employed plumbers.Not “an SMB solution”. Nobody types that. An engine can only match you to a customer if something it reads says you fit them. That’s my reasoning, not a result.
Then measure it the same way: write each customer down as the prompt they’d type, run it on a stated date on named engines, and record whether you’re named. Five labels every AI visibility number needs covers what to record.
What I will not claim
- That writing the customer into the page gets you named. I tested what changes the answer, not what changes a page’s chances.
- How the engines choose.
- Anything beyond two engines, three prompts, one evening and temporary chats. “Name three” forces a three-item list.
This is hack 3 on the search hacks page.