A prompt set is a documented list of high-intent questions buyers actually type into answer engines — used as the yardstick for whether you are named, linked, or absent. It is useful when teams need comparable measurements week over week instead of one-off screenshots.
Whilst a prompt set sounds like a keyword list, in practice vanity brand queries inflate results and miss the sale. For example, “What is [Company]?” looks fine while “best EU CNC partner for medical parts” still cites a competitor. We often recommend 5–10 real buyer prompts per market, the same wording across engines, and logging named / linked / absent before expanding the set.
Without a fixed prompt set, GEO work cannot be falsified — every demo picks a friendly question. Good sets mix category, comparison, locality, and “who is…” prompts; they stay stable long enough to show trend, then revise when the offer or market shifts. Re-run on at least two engines. Store outcomes next to URLs that should win, so gaps map to pages you can fix. Treat the set as a product artifact owned by marketing and sales, not a one-time workshop whiteboard.