Kept AI
Back

Analytics & Monitoring

Find model visibility gaps

Compare returned model results and inspect sampled answers behind the differences.

Connect your Kept MCP
Kept AIAI assistantYour workspace
Open Kept AI

Command

For my supplied host and days of 7, 30, or 90, call list_projects and verify access first. Read get_overview({host,days}) and list_prompts({host,status:"active",limit:100}); follow nextCursor with cursor where present. Compare only model breakdowns actually returned. Select prompt IDs associated with a visible difference and call get_prompt_runs({host,prompt_id,days}). Summarize brand mentions, answer framing, and citation evidence by model when present. Do not invent model scores or infer a cause from unequal samples. Produce a model comparison table, three testable explanations, and a short investigation plan; mark unavailable model detail unknown.

What you need

  • Supply a visible project host and a 7, 30, or 90 day window.
  • Model comparisons require model labels and breakdowns in the returned data.
Read the setup guide

Separate model results from sampling gaps

Compare the available model breakdowns before examining individual answers. Unequal sample coverage and missing labels stay visible in the table, helping the team distinguish an observed visibility difference from a comparison the available data cannot reliably support yet.

Turn differences into investigation questions

Use sampled answers to explain what differs in wording, brand inclusion, or citations. The deliverable pairs each hypothesis with a concrete follow-up check and its evidence, rather than presenting a plausible explanation as a confirmed diagnosis of model behavior.