Prompt panel
The panel is the measuring instrument. Its value comes entirely from not changing.
One question on the panel“Best Korean sunscreen for oily skin?”
- ChatGPTnamed
- Perplexitynot named
- Claudenamed
- Gemininot named
- AI Overviewsnot named
2 of 5 — one question’s contribution to a named rate measured over the set you fix.
A panel’s size is an output of the first week rather than a number SearchD brings to it, because a freight marketplace and a sunscreen brand do not have the same number of questions worth buying.
| Stage | What happens |
|---|---|
| Assemble candidates | Around fifty |
| Track them | Five consecutive days |
| Read the week with the client | The client decides which are kept, and which few the work concentrates on |
How that session runs is on the method page.
Candidates are built from the brand’s own language where we can get it — search console query strings, reviews, support tickets, and how people phrase things in category forums — plus the fan-out queries engines generate for themselves. Not a keyword tool: assistants are asked in sentences.
Brand questions are kept to the smallest slice on purpose. Being named when the buyer already typed your name tells you very little. The commercially interesting number is how often you appear when they did not.
The panel is step one of six in what ChatGPT searches before it names a brand, because it is the only step that cannot be done after the fact.
Related: named rate.