list_geo_audit_responses

list_geo_audit_responses MCP tool: every engine answer of a GEO Audit, with its text, brand and competitor positions, cited sources and web searches.

Updated 2026-11-06

list_geo_audit_responses returns the answers of a GEO Audit one by one. Each row is one question asked to one AI engine, for one persona and one run. It carries the answer text, whether the audited brand is mentioned and at which position, the competitors cited with their positions, the sources the engine cited and the web searches it ran. It is the raw material behind the metrics of get_geo_audit, the same data the shared report shows answer by answer.

When to use

Call it when an agent needs to quote what an engine actually said, check why a brand is missing from an answer, or rebuild its own report from the source answers. get_geo_audit gives the aggregated picture (visibility, questions played, top domains, recommendations); this tool gives the evidence. It works on a completed audit, on a running one (the answers received so far), and on a bulk campaign target.

Input

Field Type Default Description
auditId string (CUID) - The GEO Audit id returned by list_geo_audits, or a bulk campaign target's id. Returns audit_not_found if it does not belong to your workspace.
limit integer (1-50) 10 Answers per page. Answer texts are long, so keep pages small.
cursor string - Opaque cursor from a previous page's pageInfo.nextCursor.
maxResponseChars integer (min 1) 3000 Cap on each answer's text. A longer answer is cut and flagged with responseTruncated.
filters.llm enum - One engine only, by its key (for example CHATGPT). get_audit_options lists the keys.

GEO Audits are workspace-scoped, so the tool takes no projectId. A key restricted to specific projects cannot read audits.

Response

{
  "data": [
    {
      "id": "cmw5ewvxp006dzko7tekni1rx",
      "question": "Which HR consulting firm should an SME choose?",
      "persona": null,
      "llm": "CHATGPT",
      "brandMentioned": true,
      "brandPosition": 4,
      "sentiment": "positive",
      "competitors": [
        { "name": "Lumen Partners", "position": 1, "context": "**Lumen Partners** (lumenpartners.com): often cited in comparisons.", "url": "lumenpartners.com" }
      ],
      "citations": ["https://www.lumenpartners.com/", "https://www.trustpilot.com/comparatif"],
      "webSearchQueries": ["Which HR consulting firm should an SME choose?"],
      "createdAt": "2026-11-06T20:13:44.749Z",
      "response": "Here are the most recommended options...",
      "responseLength": 574,
      "responseTruncated": false
    }
  ],
  "pageInfo": { "hasMore": true, "nextCursor": "cmw5ewvxp006dzko7tekni1rx", "totalCount": 120 },
  "summary": { "hiddenLlms": [] }
}

Field notes:

  • Rows are ordered by question, then engine. persona is { id, name } when the audit ran with personas, null otherwise.
  • brandPosition is the brand's rank among the brands named in the answer, null when it is not mentioned. sentiment is positive, neutral or negative when mentioned.
  • competitors lists the other brands named in the answer, with their position and the sentence that names them.
  • citations are the URLs the engine cited. webSearchQueries are the searches it ran before answering (empty for engines that do not expose them).
  • responseLength is the full length of the answer, so an agent knows how much was cut.
  • summary.hiddenLlms lists engines left out because too few of their answers were usable. Their answers never appear, even with filters.llm.
  • For a bulk campaign target, the answers come from the campaign's shared scan, and the brand fields are those of that target.

Not found

{ "success": false, "error": "audit_not_found", "message": "No GEO audit with this id exists in your workspace. Verify the auditId returned by list_geo_audits." }

Tips and patterns

  • Read the summary first. Call get_geo_audit for the aggregated metrics, then this tool only for the answers you need to quote or inspect.
  • Narrow by engine. filters.llm keeps one engine's answers and makes totalCount the number of answers for that engine.
  • Raise the cap only when needed. The default 3000 characters covers most answers. Pass a larger maxResponseChars with a smaller limit to read a long answer in full.
  • Same data as the REST API. GET /api/v1/audits/{id}/responses returns the same rows, with the full answer text.

Related tools

You can see where you stand today.

Free trial. Start tracking your AI visibility, no credit card.