Prompt corpus & visibility baseline

This is the project that makes every future “gain” judgeable. A journey-staged prompt corpus, run across the surfaces you scoped, produces both the baseline and the noise floor — the natural week-to-week churn that a real citation lift has to beat. Without it, the program is optimizing against anecdotes.

Why this, mechanically. ChatGPT’s Instant and Thinking modes cite largely different domains, so a corpus that samples only the default mode sees a fraction of the picture. Building the prompt set per market and product line, across funnel stages (problem → exploration → comparison → validation → selection) and both reasoning modes, is what makes share of voice a defensible number rather than a lucky snapshot.

Deliverable. A journey-staged prompt corpus per priority market for each archetype, run monthly across the scoped surfaces (both ChatGPT modes included), reporting rolling 90-day share of voice and the measured noise floor.

Finish line. The baseline exists, the noise floor is quantified, and someone reads the report on a schedule.

Procedure: build and maintain the prompt library and set up AI visibility measurement. Verify against the measurement checklist; read the numbers with reading AI visibility metrics.