Skip to main content
Semantic search over the brand’s tracked prompts in Citable’s Prompt Intelligence system. Returns matching prompts with journey stage (awareness_discovery / consideration_evaluation / purchase_intent / support_usage), strategic value, topic cluster, brand scope, persona match, platforms tracked (ChatGPT/Gemini/Claude/Perplexity/Grok), and latest visibility/citation/sentiment scores. Many recent prompts have no stage yet, so a stage-filtered list is not the whole set. Use this when the user asks:
  • “Do I track prompts about topic X?”
  • “Which of my tracked questions about a topic do AI answers leave me out of?”
  • “Show me prompts tied to persona Y.”
  • “Which prompts mention competitor Z?”
Skip when the user wants a summary (use get_brand_visibility) or action items (use get_geo_priorities). Response shape: { query, total, results: [{ id, promptText, funnelStage, strategicValue, topicCluster, brandScope, personaMatch, platforms[], visibilityScore, citationScore, brandSentimentScore, isNew, trend }] }. The “trend” field is a tested verdict on the last 30 days vs the 30 before, not a raw difference. Branch on trend.status:
  • “moved” — the change is larger than sampling noise (two-proportion test, 95%). Safe to act on. trend.pp is the size in percentage POINTS, trend.direction the sign.
  • “inconclusive” — measured, but not distinguishable from noise. Do NOT describe it as rising, falling, or stable; trend.reason says why (“within-noise”, “small-sample”, “measurement-changed” = the engines scanned differed, so the two rates aren’t comparable).
  • “unmeasured” — no runs in one or both windows. Nothing to say. trend.summary is a ready one-line phrasing. Most prompts are “inconclusive” at current scan volumes — that is the honest answer, not a gap.
Example: calling with { query: “AI marketing automation”, funnelStage: “consideration_evaluation”, limit: 10 } returns up to 10 consideration_evaluation prompts semantically matching “AI marketing automation”.

​
Inputs