What a prompt set is and how to build one

Where the questions come from

Sales calls and notes, verbatim where possible. Support and intake conversations. Search queries from Search Console and the paid search terms that produced inquiries. Forums, review platforms and association channels. Competitor FAQ pages. The sources are deliberately diverse so the set reflects what buyers ask rather than what the organization assumes they ask.

Phrasing and clustering

Each question is rewritten as a buyer would type or say it to an engine, preserving their vocabulary rather than the organization’s. Questions are tagged by decision stage (understanding, options, selection, action) and clustered by intent, service and audience. Follow-up questions are included for conversational surfaces such as AI Mode.

Versioning and cadence

The set is versioned. Buyer language changes and the set follows, but every change is recorded so that a trend break can be attributed to the set rather than to the engines. Prompts run on a schedule across the named engines, repeatedly, because engines are non-deterministic; results are frequencies, not single observations.

What it feeds

The prompt set drives share-of-voice measurement, citation tracking, competitor benchmarking and the editorial plan, since each priority question maps to a page type. The buyer-question inventory worksheet walks through the build. See prompt set design.