Prompt Gaps

Overview

Prompt Gaps finds the queries you're losing: prompts where AI engines mention your competitors but not you. It is computed entirely from your own snapshot data — competitors_mentioned vs brand_mentioned per answer — so every gap is a measured fact, not a guess.

How the gap score works

For each tracked prompt over the analysis window — your last three capture cycles, ending on your latest capture (three months for a project captured monthly, three weeks for one captured weekly):

gap score = competitor mention rate (Wilson 95% lower bound) × (1 − brand mention rate) × 100

A high score means a competitor appears in essentially every answer while you appear in none — and there are enough captures to be confident it's real, not a one-off. Two safeguards keep the ranking honest: a prompt needs at least 5 snapshots before any gap is scored, and the competitor's mention rate uses the Wilson score-interval lower bound rather than the raw rate, so a competitor seen 5 of 5 times scores lower (~57) than one seen 50 of 50 times (~93) instead of both pinning at 100. Gaps are ranked biggest-first.

A question with fewer than 5 answers in the window is left out, not counted as "no gap". Every run says so in plain words: how many questions were scored, how many were skipped for sample size, and the median answers per question. If a run scores nothing, the page says that no question had enough answers — which is a different thing from having no gaps. More frequent captures, or more engines, fill it in.

The "what to write" brief

Every gap expands into a deterministic content brief — generated instantly from the measured numbers, no LLM:

• Working title shaped to the query
• Rationale with the actual mention counts and window
• Outline adapted to the prompt's form:
  • comparison ("vs") prompts → verdict-first opening + comparison table (the most-extracted format) + honest competitor-strengths section
  • list ("best/top") prompts → the list itself on the first screen + per-option self-contained summaries + ranking methodology
  • question prompts → a complete 40–80-word answer first (44% of AI citations come from a page's first 30%), then question-format sections
• Format notes carrying the proven citation levers: 3+ outbound citations (+29%), statistics (+34%), a named author byline

Copy the brief with one click and hand it to whoever writes the content.

Workflow

1. Click Analyze gaps (needs recent snapshots — gaps are computed from real AI answers).
2. Work the list top-down; the biggest gap is the most valuable page you haven't written.
3. The top gaps also appear in the Suggested inbox on your Action list, beside everything else the platform found.
4. After publishing, re-run the analysis in a few weeks — a closed gap means your brand now appears for that prompt.