A discovery prompt elicits a list or recommendation of providers, as opposed to a brand prompt (about one named brand) or an informational prompt (a how/what question with no recommendation). It is the unit of analysis that visibility metrics are computed over.
Rhinegold's discipline: the discovery-prompt set is the measurement instrument, and its design decides whether any downstream number means anything. A set must cover the real buyer questions, stay neutrally phrased, and remain stable over time. A biased or drifting prompt set produces biased, drifting Mention Rate, Citation Rate and Brand Recommendation Share — garbage in, garbage out.
It is the foundation under every visibility metric: get the prompt set right, or every number built on it is suspect.
Prompt sets drift as products and providers change; small sets are volatile; and phrasing alters results. A set has to be versioned and held stable to support a trend, and absolute levels still move with provider updates.
Across a stable, versioned set, relative movement stays reliable even when absolute levels shift with provider updates — so trend reads survive the volatility that point estimates do not.
Against a brand prompt, which asks about one named brand. Against an informational prompt, which seeks an explanation, not a recommendation. And against the metrics — Mention Rate, Share of Voice, Brand Recommendation Share — which are computed over the discovery-prompt set, not the set itself.
There is no standard taxonomy of prompt types or of prompt-set design across tools, so what counts as a "discovery" prompt — and how a set is built — varies widely.