Monitor Evidence: Cited Sources, Answers and Result Exports

Every number in a monitor report comes from scored answers you can open. This page covers the read endpoints that show them: which sources are cited, how citations move over time, the answer behind one score, one prompt in detail, answers per location, and the full export. They read stored results only, so they cost nothing.

Endpoint	Answers
`GET /v1/monitors/{id}/sources`	Which domains or pages are cited, with every row reachable by cursor.
`GET /v1/monitors/{id}/citations`	Citation totals and daily series for the top domains and pages.
`GET /v1/monitors/{id}/answers/{taskId}`	The stored answer behind one score.
`GET /v1/monitors/{id}/prompt`	One prompt’s daily series, sources and recent answers.
`GET /v1/monitors/{id}/grid`	Answers per point of a monitor’s location or grid.
`GET /v1/monitors/{id}/results`	The scored answers themselves, as JSON or CSV.

Which sources are cited

GET /v1/monitors/{id}/sources (reference) ranks the domains or pages the engines cited, using the same window and filters as the report. It counts distinct answers, not raw citations: a page cited three times in one answer counts once.

curl "https://api.querying.ai/v1/monitors/$MONITOR_ID/sources?days=30&groupBy=domain&limit=2" \
  -H "Authorization: Bearer $QUERYING_API_KEY"

Parameter	Description
`groupBy`	`domain` (default): the host without `www.`, port or credentials. `page`: the full URL.
`limit`	1–100, default 20.
`cursor`	The previous response’s `nextCursor`, unchanged.
`days`, `since`, `until`, `engine`, `prompt`	As on the report.

For example:

{
  "success": true,
  "data": {
    "results": [
      {
        "key": "reviews.example",
        "domain": "reviews.example",
        "url": null,
        "label": "Best wireless earbuds of 2026",
        "citations": 11,
        "prompts": 2,
        "own": false,
        "evidenceTaskId": "6f1c2a9e-3b4d-4e5f-8a7b-1c2d3e4f5a61"
      },
      {
        "key": "rtings.example",
        "domain": "rtings.example",
        "url": null,
        "label": "Earbuds reviews",
        "citations": 7,
        "prompts": 1,
        "own": false,
        "evidenceTaskId": "6f1c2a9e-3b4d-4e5f-8a7b-1c2d3e4f5a62"
      }
    ],
    "nextCursor": "eyJrZXkiOiJydGluZ3MuZXhhbXBsZSIsImNpdGF0aW9ucyI6N30",
    "groupBy": "domain",
    "window": { "since": "2026-09-19T07:00:00.000Z", "until": "2026-10-19T07:00:00.000Z" },
    "filters": { "engine": null, "prompt": null },
    "counting": "distinct_answers",
    "ownership": "current_monitor_domains"
  }
}

Field	Meaning
`citations`	Answers that cited this domain or page.
`prompts`	Distinct prompts whose answers cited it.
`own`	Whether it is one of the monitor’s current `domains` or a subdomain of one (`ownership: "current_monitor_domains"`).
`url`	The page URL with `groupBy=page`; `null` for domains.
`label`	The citation’s label as the engine gave it, if any.
`evidenceTaskId`	One answer that cited it. Open it with the answers endpoint.

Rows are ordered by citations, then key. nextCursor is null on the last page; there is no top-N cut, so every cited domain or page in the window is reachable. Treat the cursor as opaque.

Citations over time

GET /v1/monitors/{id}/citations (reference) returns the data behind the dashboard’s citation charts: totals for the window, a daily series, and the top 20 domains and pages, each with its own daily series.

curl "https://api.querying.ai/v1/monitors/$MONITOR_ID/citations?days=30&kind=owned" \
  -H "Authorization: Bearer $QUERYING_API_KEY"

Parameter	Description
`days`	`7`, `30` (default) or `90`.
`since`, `until`	A fixed window, as on the report. When set, `days` is only echoed back as `windowDays`.
`engine`	One engine, or `all` (default).
`prompt`	One prompt, by exact text.
`kind`	`all` (default), `owned`, `editorial`, `review_media`, `pr_wire`, `institution`, `reviews`, `commerce`, `social`, `blog`, `community`, `video`, `brand`, `content` or `other`.
`q`	Keeps domains whose name, or pages whose URL, contains this text. Up to 200 characters.
`offset`	0–10,000, for paging the domain and page lists 20 at a time.

Here a citation is one page in one answer, so citations counts cited pages, not answers. /sources counts distinct answers, so its numbers can differ.

Field	Contents
`totals`	`answers` (scored answers), `citedAnswers`, `citations` and `ownedCitations` for the window.
`days`	One entry per UTC day with answers: `day`, `answers`, `citations`, `ownedCitations`. A day with answers and no citations has `citations: 0`; a day with no answers is absent.
`domains`, `pages`	Up to 20 rows from `offset`: `key`, `domain`, `label`, `kind`, `owned`, `citations`, `answers`, `prompts`, and `daily`, a list of `{ day, citations }` for days with citations.
`types`	Per `kind`: `citations` and distinct `domains`.
`pagination`	`offset`, `limit` (20), `totalDomains`, `totalPages`.
`windowDays`, `engine`, `prompt`, `since`, `until`	The scope that was read.

kind is owned for your domains; other kinds come from a fixed list of known publishers, and anything unknown is other. A source’s share is its citations divided by totals.citations. kind, q and offset narrow only domains, pages and pagination; totals, days and types always cover every source in the window.

Read one answer

GET /v1/monitors/{id}/answers/{taskId} (reference) returns the stored answer behind one score, with what it was matched against, so a mention can be checked rather than trusted. Task ids come from evidenceTaskId, mentionTaskId and latestTaskId in a report, from /sources, and from exported rows.

curl "https://api.querying.ai/v1/monitors/$MONITOR_ID/answers/6f1c2a9e-3b4d-4e5f-8a7b-1c2d3e4f5a61" \
  -H "Authorization: Bearer $QUERYING_API_KEY"

For example:

{
  "success": true,
  "data": {
    "answer": {
      "taskId": "6f1c2a9e-3b4d-4e5f-8a7b-1c2d3e4f5a61",
      "engine": "CHATGPT",
      "prompt": "best wireless earbuds for commuting",
      "mentioned": true,
      "cited": true,
      "position": 81,
      "ranAt": "2026-10-17T09:01:48.311Z",
      "answerText": "For commuting, look at the **Northwind Audio Pro** for noise cancelling or the **Acme Audio Air** for battery life.",
      "sources": [
        { "url": "https://reviews.example/best-earbuds", "label": "Best wireless earbuds of 2026", "position": 1 },
        { "url": "https://acme-audio.example/air", "label": "Acme Audio Air", "position": 2 }
      ],
      "competitorHits": ["Northwind Audio"],
      "scoringContext": {
        "aliases": ["Acme Audio", "Acme"],
        "domains": ["acme-audio.example"],
        "competitors": [
          { "name": "Northwind Audio", "aliases": ["Northwind Audio"], "source": "user" },
          { "name": "Contoso Sound", "aliases": ["Contoso Sound", "Contoso"], "source": "user" }
        ],
        "country": "US",
        "version": "3f9a6c1e8b2d4f7a9c0e1b3d5f7a9c1e3b5d7f9a1c3e5b7d9f1a3c5e7b9d1f3a",
        "scoredAt": "2026-10-17T09:01:48.305Z",
        "answerPresent": true,
        "evidenceTruncated": false
      },
      "analysis": null
    },
    "display": {
      "markdown": "For commuting, look at the **Northwind Audio Pro** for noise cancelling or the **Acme Audio Air** for battery life.",
      "text": "For commuting, look at the Northwind Audio Pro for noise cancelling or the Acme Audio Air for battery life."
    },
    "mode": "BRAND",
    "aliases": ["Acme Audio", "Acme"],
    "competitors": [
      { "name": "Northwind Audio", "aliases": ["Northwind Audio"], "source": "user" },
      { "name": "Contoso Sound", "aliases": ["Contoso Sound", "Contoso"], "source": "user" }
    ],
    "scoringKnown": true,
    "evidenceTruncated": false
  }
}

Field	Meaning
`answer`	The scored answer, as stored.
`answer.answerText`	The answer, as markdown when the engine gave it, cut at 8,000 characters with a trailing `…`. `null` when the engine returned no text.
`answer.sources`	The cited pages in the engine’s order, deduplicated by URL, each with its `label` and 1-based `position`.
`answer.competitorHits`	The competitors named in the answer.
`answer.position`	Character offset of the first match of your brand, or `null`.
`answer.scoringContext`	The definition the answer was scored with: `aliases`, `domains`, `competitors`, `country`, its `version` hash, `scoredAt`, `answerPresent` and `evidenceTruncated`. `null` for answers scored before it was stored.
`answer.analysis`	The answer analysis result when the monitor has `analysis: true`, otherwise `null`.
`display`	Display fields of the original task response, such as `markdown`, `text`, `sources`, `searchQueries`, layout, places and products, depending on the engine. `null` when the original task is no longer stored.
`mode`	The monitor’s mode. On a category monitor `aliases` is empty and `competitors` is the list of tracked brands.
`aliases`, `competitors`	What this answer was matched against: from `scoringContext`, or the monitor’s current values when it is `null`.
`scoringKnown`	`false` when the answer predates stored scoring context; `aliases` and `competitors` are then the monitor’s current values.
`evidenceTruncated`	Whether `answerText` was cut. `null` when `scoringKnown` is `false`.

A task id that is not a scored answer of this monitor returns 404 NOT_FOUND.

One prompt in detail

GET /v1/monitors/{id}/prompt (reference) returns one prompt’s data. prompt (exact text) is required, and works for prompts you have since removed from the monitor; days is 7, 30 (default) or 90.

curl -G "https://api.querying.ai/v1/monitors/$MONITOR_ID/prompt" \
  --data-urlencode "prompt=best wireless earbuds for commuting" \
  --data-urlencode "days=30" \
  -H "Authorization: Bearer $QUERYING_API_KEY"

Field	Contents
`monitor`	`id`, `name`, `engines` and `aliases`.
`prompt`, `windowDays`	What was read.
`series`	Per UTC day and engine: `day`, `engine`, `runs`, `mentioned`, `cited`.
`topDomains`	Up to 12 cited domains: `domain`, `citations`, `prompts`, `own`.
`topPages`	Up to 20 cited pages: `url`, `domain`, `label`, `citations`, `prompts`, `own`.
`results`	Up to 100 latest scored answers: `taskId`, `engine`, `prompt`, `mentioned`, `cited`, `position`, `ranAt`, `competitorHits`.

Here citations counts cited pages across answers, not distinct answers. For a full, paginated source list for one prompt, use /sources?prompt=....

Answers by location

GET /v1/monitors/{id}/grid (reference) returns a monitor’s answers per point of its current location or grid. Optional engine (one of the monitor’s engines) and prompt narrow it; days is 7, 30 (default) or 90. The response has location, grid, engine, prompt, windowDays and points. Each point has row, col, location, runs, mentioned, cited, topCompetitor ({ name, mentions } or null) and latest ({ taskId, engine, mentioned, ranAt } of the newest answer, or null). Answers are matched to points by their exact coordinates, so answers from an earlier location or grid of the same monitor are left out. A monitor without a location or grid returns no points.

Export scored answers

GET /v1/monitors/{id}/results (reference) returns the scored answers themselves, for a warehouse load, a spreadsheet or a deck. To deliver every new answer to cloud storage on a schedule instead, see Result exports.

curl "https://api.querying.ai/v1/monitors/$MONITOR_ID/results?since=2026-10-01T00:00:00Z&until=2026-10-18T00:00:00Z&mentioned=false" \
  -H "Authorization: Bearer $QUERYING_API_KEY"

Parameter	Description
`since`, `until`	Half-open window. Default: the 30 days ending now. ISO instants with a timezone; a bare `YYYY-MM-DD` is also accepted here and read as midnight UTC. There is no maximum span.
`engine`, `prompt`	As on the report.
`mentioned`, `cited`	`true` or `false`. `mentioned=false` is the gap view.
`competitor`	Keeps answers whose stored competitor hits include this exact name, 1–80 characters. It matches the name as it was stored when each answer was scored and is not checked against the monitor’s current competitors: a name the monitor never tracked returns no rows, not an error.
`includeEvidence`	`true` adds `answerText`, `sources` and `scoringContext` to every row.
`limit`	1–10,000, default 10,000. With `includeEvidence=true`: 1–100, default 100.
`format`	`json` (default) or `csv`.
`cursor`	The previous page’s `nextCursor`.

The JSON response, for example:

{
  "success": true,
  "data": {
    "monitor": { "id": "d39b8de0-b159-4229-bf58-7064e5369db1", "name": "Earbud brand tracking" },
    "since": "2026-10-01T00:00:00.000Z",
    "until": "2026-10-18T00:00:00.000Z",
    "filters": { "engine": null, "prompt": null, "mentioned": false, "cited": null, "competitor": null },
    "truncated": false,
    "nextCursor": null,
    "results": [
      {
        "id": "0b6e4a52-7c1d-4f8e-9a3b-2d5c6e7f8a91",
        "taskId": "6f1c2a9e-3b4d-4e5f-8a7b-1c2d3e4f5a63",
        "engine": "CHATGPT",
        "prompt": "are cheap earbuds worth it",
        "mentioned": false,
        "cited": false,
        "position": null,
        "ranAt": "2026-10-17T09:02:57.204118Z",
        "competitorHits": ["Northwind Audio"]
      }
    ]
  }
}

Rows are newest first. Pagination is a keyset over ranAt (microsecond precision) and row id, so rows that share a timestamp are never skipped or repeated. For a repeatable extraction, send an explicit since and until, keep them fixed, and follow nextCursor until it is null. Changing the window between pages can skip or repeat rows. Treat the cursor as opaque.

CSV

With format=csv the response is text/csv, UTF-8 with a byte order mark so spreadsheet apps open it correctly. A CSV body cannot carry pagination, so it moves to headers:

Header	Meaning
`x-next-cursor`	The cursor for the next page; empty on the last page.
`x-result-truncated`	`true` when more rows matched than this page returned.
`x-result-since`, `x-result-until`	The resolved window, so a resumed export can pin it.

Columns are ran_at, monitor, engine, prompt, mentioned, cited, position, competitors, task_id, plus answer_text, sources, scoring_context with includeEvidence=true. competitors joins names with |; sources and scoring_context hold JSON text. A cell that starts with =, +, -, @, a tab or a carriage return is prefixed with ', so engine text cannot become a spreadsheet formula.

curl -D headers.txt -o page1.csv \
  "https://api.querying.ai/v1/monitors/$MONITOR_ID/results?since=2026-10-01T00:00:00Z&until=2026-10-18T00:00:00Z&format=csv&includeEvidence=true" \
  -H "Authorization: Bearer $QUERYING_API_KEY"

Errors and limits

Limit	Value
Report and source window	90 days
`/sources` page	1–100 rows, default 20
`/citations` lists	20 rows per page; `offset` up to 10,000
`/results` page	10,000 rows; 100 with `includeEvidence=true`
Stored answer text	8,000 characters
`/prompt`	12 domains, 20 pages, 100 answers

Invalid parameters return 400 VALIDATION_ERROR naming the problem, including a cursor that was not returned by the previous page. A monitor of another account, or an unknown task id, returns 404 NOT_FOUND. See Monitors for the shared codes.

Get started

Core concepts

Answer engines

Search and web data

Request options

Research

GEO monitoring

Account and data

Monitor Evidence: Cited Sources, Answers and Result Exports

Which sources are cited

Citations over time

Read one answer

One prompt in detail

Answers by location

Export scored answers

CSV

Errors and limits

Next steps

Events and alerts

Result exports

​Which sources are cited

​Citations over time

​Read one answer

​One prompt in detail

​Answers by location

​Export scored answers

​CSV

​Errors and limits

​Next steps

Events and alerts

Result exports

Which sources are cited

Citations over time

Read one answer

One prompt in detail

Answers by location

Export scored answers

CSV

Errors and limits

Next steps