AI Sources

REST endpoints for AI Sources. Base URL https://api.competlab.com; every request needs a CL-API-Key header (see Authentication). Responses wrap in { item } or { items }; errors in { error }.

Get latest AI sources data

GET/v1/projects/{projectId}/ai-sources

Get the latest AI Sources data for a project: for each engine we ask — Perplexity and Google AI Overviews — which companies it named in answer to the project’s 8 buying questions, which pages it RETRIEVED to answer, and which of the pages more than one engine retrieved name the customer’s competitors and not the customer. Every number is read off the stored summary of the latest published check, the same document the app shows. Two sentences this data cannot support, and no field here claims: a CITATION count — an engine hands back the pages it retrieved while answering and does not say which it leaned on — and a POOLED page number — the engines read different pages, so page counts are per engine and are never added together. The one cross-engine object is the core, the hosts at least 2 engines retrieved (summary.overlap, summary.coreHosts); answers may pool as a vote on the leaderboard (summary.brands), and every such row says which engines back it. Counts are counts, never rates: report figures as n of N answers and never as a percentage or a share — the question set is small by design, and a share computed from it is false precision. Every count names its universe on the same object. An engine the check did not ask is ABSENT from every per-engine record, and absent means not measured, never zero; an engine present with engineDataAvailable produced nothing usable and carries no count. The question matrix has three cell states, never two: answered, no_answer_shown (the engine was read and showed nothing — measured, not an answer, not a failure) and not_measured (we could not read it — not a zero). A page we could not read is listed under its host with its reason and is never counted as a page the customer is absent from; only a core host whose status is missing supports ‘get onto this page’. actionHint.text on each core host and the sentences under summary.limits.sentences are payload: render them verbatim. Set includeAnswers=true to also get the engines’ raw answers on that check — the answer text, the companies read out of it in order of first mention, and the pages each engine retrieved with the passage it handed back — narrowed by engine= or promptIndex=. The block is large; the prose in it is the engine’s, not CompetLab’s. No filter changes a number under summary. When the most recent check was abandoned because no engine produced a usable answer, latestCheckDataAvailable is present and everything else describes an earlier check. A project with no published check yet answers 404 no_data_available — not a missing project and not a failed measurement, just nothing measured here so far; the project itself is fine and get_project still describes it. Distinguish it from project_not_found, which means the id is wrong.

Returns · { item }

Path parameters

Name	Type	Description
`projectId`	`string`	Project ID

Query parameters

Name	Type	Description
`includeAnswers`	`boolean`	Set true to include what the engines actually said — every question sent, the answer text, the companies read out of it in order of first mention, and the pages the engine RETRIEVED to write it, with the passage it handed back for each. Off by default because the block is large: up to 8 answers per engine, each carrying the engine’s full retrieved page list, which for a searching engine can run to dozens of pages with a passage each. Narrow with `engine=` or `promptIndex=` rather than fetching everything. The answer text is the engine’s wording, not CompetLab’s assessment; the pages are retrieved, never cited. Default `false`.
`engine`	`string`	Return only this engine’s answers. Requires `includeAnswers=true`. Narrows `answers`, `unansweredQueries` and `noAnswerShown`; changes nothing under `summary` and does not narrow `engineStatus`. Values: `perplexity`, `google_ai_overviews`.
`promptIndex`	`number`	Return only the answers for this question, across every engine. Requires `includeAnswers=true`. Zero-based: the question’s position in the check’s question list, matching `promptIndex` on each answer.

Request


curl "https://api.competlab.com/v1/projects/507f1f77bcf86cd799439011/ai-sources?includeAnswers=false&engine=perplexity&promptIndex=0" \
  -H "CL-API-Key: YOUR_COMPETLAB_API_KEY"

Response 200 OK

Example


{
  "item": {
    "lastUpdatedAt": "2026-09-04T10:00:00.000Z",
    "summary": {
      "customerDomain": "mycompany.com",
      "verdict": "missing_from_most_core_hosts",
      "window": {
        "checkIds": [
          "507f1f77bcf86cd799439012",
          "507f1f77bcf86cd799439011"
        ],
        "since": "2026-08-01T10:00:00.000Z",
        "perEngine": {
          "perplexity": {
            "checkIds": [
              "507f1f77bcf86cd799439012"
            ]
          },
          "google_ai_overviews": {
            "checkIds": [
              "507f1f77bcf86cd799439012"
            ]
          }
        }
      },
      "perEngine": {
        "perplexity": {
          "engineDataAvailable": {
            "available": false,
            "reason": "no_usable_answer"
          },
          "answersAsked": 8,
          "answersReceived": 7,
          "answersAbsent": 1,
          "answersUnmeasured": 0,
          "answersWithoutRetrieval": 1,
          "answersNamingCustomer": 3,
          "answersRetrievingOwnPage": 4,
          "pagesRetrieved": 131,
          "hostsRetrieved": 58,
          "hostsOnlyThisEngine": 33,
          "engineCoreSize": 35,
          "pagesRead": 118,
          "pagesUnreadable": 9,
          "pagesNotFetched": 4,
          "ownPagesRetrieved": 6,
          "independentPagesNamingCustomer": {
            "floor": 4,
            "ceiling": 6
          },
          "topHosts": [
            {
              "host": "g2.com",
              "pages": 3,
              "promptIds": [],
              "alsoRetrievedBy": [
                "google_ai_overviews"
              ]
            }
          ]
        },
        "google_ai_overviews": {
          "engineDataAvailable": {
            "available": false,
            "reason": "no_usable_answer"
          },
          "answersAsked": 8,
          "answersReceived": 7,
          "answersAbsent": 1,
          "answersUnmeasured": 0,
          "answersWithoutRetrieval": 1,
          "answersNamingCustomer": 3,
          "answersRetrievingOwnPage": 4,
          "pagesRetrieved": 131,
          "hostsRetrieved": 58,
          "hostsOnlyThisEngine": 33,
          "engineCoreSize": 35,
          "pagesRead": 118,
          "pagesUnreadable": 9,
          "pagesNotFetched": 4,
          "ownPagesRetrieved": 6,
          "independentPagesNamingCustomer": {
            "floor": 4,
            "ceiling": 6
          },
          "topHosts": [
            {
              "host": "g2.com",
              "pages": 3,
              "promptIds": [],
              "alsoRetrievedBy": [
                "google_ai_overviews"
              ]
            }
          ]
        }
      },
      "matrix": [
        {
          "promptId": "7f3c1a2e-9d44-4c8b-9f0e-2b6a5c9d7e10",
          "promptText": "Which email warm-up tools are safe to use with a Google Workspace domain?",
          "intent": "safety",
          "cells": {
            "perplexity": {
              "state": "answered",
              "reason": "no_ai_overview_shown",
              "named": false,
              "companiesNamed": 6,
              "webSearch": {},
              "pagesRetrieved": 14,
              "pagesRead": 12,
              "pagesUnreadable": 1,
              "pagesNotFetched": 1,
              "pagesNamingCustomer": {},
              "ownPagesRetrieved": 1
            },
            "google_ai_overviews": {
              "state": "answered",
              "reason": "no_ai_overview_shown",
              "named": false,
              "companiesNamed": 6,
              "webSearch": {},
              "pagesRetrieved": 14,
              "pagesRead": 12,
              "pagesUnreadable": 1,
              "pagesNotFetched": 1,
              "pagesNamingCustomer": {},
              "ownPagesRetrieved": 1
            }
          }
        }
      ],
      "brands": [
        {
          "domain": "competitor.com",
          "name": "Competitor Inc",
          "isOwn": false,
          "origin": "tracked_competitor",
          "ambiguous": false,
          "rankByPresence": 2,
          "answers": {
            "answersNaming": 5,
            "answersReceived": 16,
            "presence": 31,
            "presenceLow": 14,
            "presenceHigh": 56
          },
          "answersRetrievingOwnSite": {
            "answersNaming": 5,
            "answersReceived": 16,
            "presence": 31,
            "presenceLow": 14,
            "presenceHigh": 56
          },
          "enginesNaming": [
            "perplexity"
          ],
          "perEngine": {
            "perplexity": {
              "answers": {},
              "independentPagesNaming": {},
              "pagesRead": 118,
              "ownPagesRetrieved": 2
            },
            "google_ai_overviews": {
              "answers": {},
              "independentPagesNaming": {},
              "pagesRead": 118,
              "ownPagesRetrieved": 2
            }
          }
        }
      ],
      "overlap": {
        "hostsInCore": 25,
        "perEngine": {
          "perplexity": {
            "hostsRetrieved": 58,
            "hostsOnlyThisEngine": 33
          },
          "google_ai_overviews": {
            "hostsRetrieved": 58,
            "hostsOnlyThisEngine": 33
          }
        }
      },
      "funnel": {
        "coreHosts": 25,
        "alreadyNamingCustomer": 19,
        "unreadable": 2,
        "missing": 4,
        "missingPublishers": 3,
        "missingCompetitorOwned": 1
      },
      "coreHosts": [
        {
          "host": "g2.com",
          "status": "missing",
          "ownership": "third_party",
          "ownedBy": "competitor.com",
          "kind": "review_site",
          "kindCheckedAt": "2026-09-04",
          "actionHint": {
            "code": "pitch_publisher",
            "text": "Pitch an update to this page: it names your competitors and not you."
          },
          "engines": [
            "perplexity",
            "google_ai_overviews"
          ],
          "promptIds": [
            "value"
          ],
          "seenInChecks": {
            "perplexity": {
              "checkIds": [
                "507f1f77bcf86cd799439012"
              ]
            },
            "google_ai_overviews": {
              "checkIds": [
                "507f1f77bcf86cd799439012"
              ]
            }
          },
          "brandsNamed": [
            "competitor.com"
          ],
          "pages": [
            {
              "key": "g2.com/categories/email-warm-up",
              "url": "https://www.g2.com/categories/email-warm-up",
              "host": "g2.com",
              "title": "Best Email Warm-Up Software",
              "engines": [
                "perplexity",
                "google_ai_overviews"
              ],
              "promptIds": [],
              "ownedBy": "competitor.com",
              "readState": "read",
              "unreadableReason": "bot_protection",
              "fetchedAt": "2026-09-04T10:00:00.000Z",
              "textSource": "body",
              "namesCustomer": "not_named",
              "brandsNamed": [
                "competitor.com",
                "rival.io"
              ],
              "excerpt": "…the most reliable options this year are Competitor Inc and Rival…"
            }
          ]
        }
      ],
      "ownPageRetrievals": [
        {
          "engine": "perplexity",
          "promptId": "7f3c1a2e-9d44-4c8b-9f0e-2b6a5c9d7e10",
          "pages": [
            {
              "key": "mycompany.com/pricing",
              "url": "https://mycompany.com/pricing",
              "title": "Pricing — My Company"
            }
          ],
          "companiesNamed": [
            "competitor.com",
            "rival.io"
          ],
          "namedCustomer": false
        }
      ],
      "pages": [
        {
          "key": "g2.com/categories/email-warm-up",
          "url": "https://www.g2.com/categories/email-warm-up",
          "host": "g2.com",
          "title": "Best Email Warm-Up Software",
          "engines": [
            "perplexity",
            "google_ai_overviews"
          ],
          "promptIds": [
            "value"
          ],
          "ownedBy": "competitor.com",
          "readState": "read",
          "unreadableReason": "bot_protection",
          "fetchedAt": "2026-09-04T10:00:00.000Z",
          "textSource": "body",
          "namesCustomer": "not_named",
          "brandsNamed": [
            "competitor.com",
            "rival.io"
          ],
          "excerpt": "…the most reliable options this year are Competitor Inc and Rival…"
        }
      ],
      "limits": {
        "pagesRetrieved": 153,
        "pagesRead": 140,
        "pagesUnreadable": 9,
        "pagesNotFetched": 4,
        "pagesTruncated": 3,
        "unreadableByReason": {
          "bot_protection": 5,
          "host_blocked_recently": 2,
          "text_too_short": 1,
          "not_found_page": 0,
          "consent_wall": 0,
          "unreachable": 0,
          "timeout": 1,
          "fetch_failed": 0,
          "unresolved": 0
        },
        "textSourceMix": {
          "body": 140,
          "snippet_only": 9,
          "none": 4
        },
        "sentences": [
          {
            "code": "sources_retrieved_not_cited",
            "text": "Retrieved, never cited: these are the pages the engine pulled back while answering; which of them it leaned on is not something the engines disclose."
          }
        ]
      }
    },
    "trackedDomains": [
      "mycompany.com",
      "competitor.com"
    ],
    "latestCheckDataAvailable": {
      "available": false,
      "reason": "no_usable_answer_from_any_engine",
      "enginesAsked": [
        "perplexity",
        "google_ai_overviews"
      ]
    },
    "answers": [
      {
        "engine": "perplexity",
        "promptIndex": 0,
        "promptId": "7f3c1a2e-9d44-4c8b-9f0e-2b6a5c9d7e10",
        "promptText": "Which email warm-up tools are safe to use with a Google Workspace domain?",
        "intent": "safety",
        "askedIn": {
          "locationName": "United States",
          "languageName": "English"
        },
        "answerText": "For a Google Workspace domain, the safest warm-up tools are Competitor Inc and Rival, both of which…",
        "companiesNamed": [
          {
            "rank": 1,
            "name": "Competitor Inc",
            "domain": "competitor.com"
          }
        ],
        "sources": [
          {
            "url": "https://www.g2.com/categories/email-warm-up",
            "domain": "g2.com",
            "title": "Best Email Warm-Up Software",
            "snippet": "Warm-up tools gradually increase sending volume so that mailbox providers…",
            "publishedAt": "2026-05-12",
            "resolvedFrom": "https://www.google.com/goto?url=…"
          }
        ],
        "searchQueries": [
          "email warm-up tools Google Workspace safe"
        ],
        "webSearch": {
          "performed": true,
          "completedSearchCount": 3,
          "reason": "NO_SEARCH_ATTEMPTED"
        }
      }
    ],
    "unansweredQueries": [
      {
        "engine": "perplexity",
        "promptIndex": 5,
        "promptId": "7f3c1a2e-9d44-4c8b-9f0e-2b6a5c9d7e10",
        "promptText": "Which email warm-up tools have a free plan?",
        "reason": "no_usable_answer"
      }
    ],
    "noAnswerShown": [
      {
        "engine": "google_ai_overviews",
        "promptIndex": 3,
        "promptId": "7f3c1a2e-9d44-4c8b-9f0e-2b6a5c9d7e10",
        "promptText": "Which email warm-up tools have a free plan?",
        "reason": "no_ai_overview_shown"
      }
    ],
    "engineStatus": {
      "perplexity": {
        "reported": true,
        "completedAt": "2026-09-04T10:00:00.000Z",
        "questionsAsked": 8,
        "answersReceived": 7,
        "answersAbsent": 1,
        "answersUnmeasured": 0
      },
      "google_ai_overviews": {
        "reported": true,
        "completedAt": "2026-09-04T10:00:00.000Z",
        "questionsAsked": 8,
        "answersReceived": 7,
        "answersAbsent": 1,
        "answersUnmeasured": 0
      }
    },
    "answersTruncated": false
  }
}

Fields

item — object

Field	Type	Description
`lastUpdatedAt`	`string`	When the data below was measured (ISO-8601): the completion time of the check the summary comes from, which is NOT necessarily the most recent cycle — see `latestCheckDataAvailable`.
`summary`	`object`	The stored summary of the latest published check, exactly as the app shows it: per engine, who was named, what was read, and which of the pages more than one engine read name the customer’s competitors and not them. See `item.summary`.
`trackedDomains`	`array<string>`	The domains the project tracks RIGHT NOW — the competitors plus the customer’s own — normalised as the summary’s domains are. Resolved against today’s roster, not the one stored with the check, so a brand on `summary.brands` whose `origin` is `named_by_engines` and whose domain is here was added to the list after the check.
`latestCheckDataAvailable`	`object`	Present ONLY when the project’s most recent check was abandoned — because no engine produced a usable answer, or because its page stage could not be closed; `reason` says which. When present, every other field in this response comes from an EARLIER check and `lastUpdatedAt` is older than the most recent check attempted. Absent when the most recent check published normally — and absent when an older check was abandoned but a later one has since published, because that failure is superseded. See `item.latestCheckDataAvailable`.
`answers`	`array<object>`	The engines’ answers for this check — one entry per question that came back with an answer, with the pages the engine RETRIEVED to write it. Present ONLY when the request set `includeAnswers=true`; absent otherwise, never an empty array standing in for ‘not requested’. Narrowed by `engine` and `promptIndex` when those are set — they narrow `unansweredQueries` and `noAnswerShown` the same way. No filter changes anything under `summary`: every number there is stored, computed over the whole check, and never recomputed for a filtered view. The pages on an answer are the pages the engine retrieved while answering, never a list of citations — the engines do not disclose which pages they leaned on. Never count pages across engines. ATTRIBUTION: the answer text and the companies read out of it are what that engine said, including about third parties CompetLab does not monitor. Attribute it to the named `engine`; do not republish it as fact. See `item.answers[]`.
`unansweredQueries`	`array<object>`	Questions this check sent that we could not read an answer to. Present ONLY when `includeAnswers=true`. Carried in their own array rather than mixed into `answers`, so that `answers.length` always means answers received. These slots are in no count on this response. Never quote this array’s length as a fraction of anything: state it as a separate fact. See `item.unansweredQueries[]`.
`noAnswerShown`	`array<object>`	Questions an engine was read for and showed no answer to — today, questions Google’s results page carried no AI Overview for. Not a failure and not an answer: in no count on this response. Separate from `unansweredQueries`, which are questions we could not read. Present ONLY when `includeAnswers=true`. Narrowed by `engine` and `promptIndex`. See `item.noAnswerShown[]`.
`engineStatus`	`object`	Per-engine reporting status for this check: whether each engine this check asked stored a result, when, and how its questions split between answered, no answer shown and unmeasured. One entry per engine the check ASKED — an engine is absent when the check did not ask it; absent means not measured, never ‘did not report’. Present ONLY when `includeAnswers=true`. NOT narrowed by the filters: it always describes the whole check. See `item.engineStatus`.
`answersTruncated`	`boolean`	True when the answers payload hit the response size cap and answers were dropped from the end of `answers`. The cut lands on a WHOLE QUESTION ROUND, never inside one: `answers` is ordered question by question across the engines, so truncation drops the highest `promptIndex` values and no engine is ever removed by truncation from a question that is still present — group the array by `promptIndex` and no round is short. Neither `unansweredQueries` nor `noAnswerShown` is ever truncated, so a question absent from `answers` is not a question no engine answered. An engine missing from a question that IS present is in one of those two arrays, not cut; an engine missing from EVERY question stored nothing for this check, which `engineStatus` reports as its three counts summing below `questionsAsked`. An answer’s page list is never partially truncated. `summary` and `engineStatus` are never affected. Narrow with `engine` or `promptIndex` to get a complete view. Present ONLY when `includeAnswers=true`.

item.summary — object

Field	Type	Description
`customerDomain`	`string`	The customer’s domain, normalised.
`verdict`	`string`	The one reading of the customer’s standing this check, decided when the summary was built and carried here as stored — a consumer never re-derives it from the numbers. Each value names a CONDITION over fields on this same object, never a rating: `recommended_nowhere` — no answer in the window named the customer (the row on `brands` with `isOwn: true` has `answers.answersNaming` of 0). `named_on_most_core_hosts` — the customer is named on at least half of the core hosts (`funnel.alreadyNamingCustomer × 2` is at least `funnel.coreHosts`, with a core to speak of): a short work list under this value is the finding, because there is little left to win. `missing_from_most_core_hosts` — otherwise: named somewhere in the window, and absent from most of the hosts more than one engine read. State the condition beside the two counts it rests on; it ranks nobody and is not a score. Values: `recommended_nowhere`, `named_on_most_core_hosts`, `missing_from_most_core_hosts`.
`window`	`object`	The checks this summary pools over, and per engine the checks that engine answered on. See `item.summary.window`.
`perEngine`	`object`	One reading per engine this check asked — Perplexity and Google AI Overviews today. Every number under a key is that engine’s alone. Keyed by engine. A key is ABSENT for an engine this check did not ask — a check keeps the engine set it was run with — and absent means not measured, never a zero. Never add the engines’ page figures together. Nothing on this surface is a citation count, and no field sums the engines’ pages. See `item.summary.perEngine`.
`matrix`	`array<object>`	One row per question the check asked, in question order, with one cell per engine. Two independent facts sit in an answered cell — whether the answer named the customer, and how many retrieved pages name them — and neither is derived from the other. See `item.summary.matrix[]`.
`brands`	`array<object>`	Who else this market names: the customer, the tracked competitors, and every company an engine named in the window. In the ORDER TO RENDER — answers naming desc, then domain — and never re-sorted by anything positional. The customer’s row is always present; read `isOwn` to find it. Ties share a `rankByPresence`. See `item.summary.brands[]`.
`overlap`	`object`	How the engines’ host sets overlap this check — the one cross-engine object. See `item.summary.overlap`.
`funnel`	`object`	The narrowing from ‘hosts more than one engine read’ to ‘hosts the customer is genuinely missing from’, in four counts that are each a filter over `coreHosts`. Show the narrowing, not only its end: on a market leader the end is small because most of the core already names them. See `item.summary.funnel`.
`coreHosts`	`array<object>`	Every core host with its standing — the work list (`missing`), the won list (`already_named`) and the hosts we could not read (`unreadable`) in ONE array, in the ORDER TO RENDER: the work list first (third-party before competitor-owned), then unreadable, then already named. `status` is the column that separates them. Never tell a customer to get onto a host whose status is not `missing`. See `item.summary.coreHosts[]`.
`ownPageRetrievals`	`array<object>`	Every answer in which an engine opened one of the customer’s own pages, with who it then recommended. Empty is measured: no engine retrieved a page of theirs this check. See `item.summary.ownPageRetrievals[]`.
`pages`	`array<object>`	Every page retrieved this check by any engine, with its read state and naming — the pages-read list. Each says which engines retrieved it; the list itself is a fetch-stage inventory, not a per-engine count, so never quote its length as an engine’s number. See `item.summary.pages[]`.
`limits`	`object`	What the numbers rest on, and the sentences to render beside them. See `item.summary.limits`.

item.latestCheckDataAvailable — object

Field	Type	Description
`available`	`boolean`	Always false. This object exists only to say the most recent check produced nothing publishable.
`reason`	`string`	`no_usable_answer_from_any_engine` — every engine the check asked came back with no usable answer, so nothing was measured and the check was abandoned. Our problem: it never locks the customer out, a new check runs automatically, and the caller may trigger one. A check where ONE engine came back empty is not this — it publishes with that engine marked unmeasured. `page_stage_failed` — the engines answered, but the stage that reads the retrieved pages could not be closed, so no summary was built and the check was abandoned. Ours, never the engines’ or the customer’s; a new check runs automatically. A page stage that merely ran out of time is not this: it closes with the pages it has, the rest recorded as not fetched, and the check publishes. Values: `no_usable_answer_from_any_engine`, `page_stage_failed`.
`enginesAsked`	`array<string>`	The engines that check asked, read off the check. Membership, never a count.

item.answers[] — each entry

Field	Type	Description
`engine`	`string`	Which engine produced this answer. Values: `perplexity`, `google_ai_overviews`.
`promptIndex`	`number`	Zero-based position of the question in this check’s question list, as the engine’s answers are stored. The value the `promptIndex` filter matches. Stable under filtering. Use `promptId` to join this row with the summary’s matrix and pages.
`promptId`	`string`	Stable identifier of the question, kept across edits of its wording.
`promptText`	`string`	The question as it was sent — one of the project’s own buying questions.
`intent`	`string`	The question’s angle. Fixed across every project; the token names the angle and nothing else. Values: `general_choice`, `problem_driven`, `price_constraint`, `company_size`, `feature`, `free_tier`, `safety`, `category_compare`.
`askedIn`	`object`	Where the engine was asked. Both engines are asked from the project’s market, so the retrieved pages are that market’s web. See `item.answers[].askedIn`.
`answerText`	`string`	The answer as the engine wrote it, verbatim — the evidence the companies below were read from. Engine prose, attributed to the engine that wrote it, never to CompetLab.
`companiesNamed`	`array<object>`	The companies the answer named, in order of first mention in `answerText`, each verified present in the text. An EMPTY list is a measurement, never missing data: the engine answered and named no company — it explained a concept and recommended nobody. A question that produced no answer is not here at all; it is in `unansweredQueries`, and a question the engine showed no answer to is in `noAnswerShown`. See `item.answers[].companiesNamed[]`.
`sources`	`array<object>`	The pages the engine RETRIEVED while it answered, distinct by URL, each with whatever the engine reports about it. Retrieved, never cited: the engine does not say which of them it leaned on, and nothing here may be called a citation. ABSENT means the engine reported no retrieval for this answer — it reported none, and has not said how it answered — never that it read nothing we could see. An EMPTY array is the measured ‘it reported retrieving nothing’. See `item.answers[].sources[]`.
`searchQueries`	`array<string>`	The search strings the engine itself issued, as it reports them. Absent where the engine reports none. Strings, not a search count: one search can fan out into several.
`webSearch`	`object`	Whether a live search ran for this answer. Absent means not recorded, never ‘did not search’. `performed: false` is the engine’s own record that no search ran, kept as it is. See `item.answers[].webSearch`.

item.unansweredQueries[] — each entry

Field	Type	Description
`engine`	`string`	Which engine this question was sent to. Values: `perplexity`, `google_ai_overviews`.
`promptIndex`	`number`	Zero-based position of the question in this check’s question list.
`promptId`	`string`	Stable identifier of the question.
`promptText`	`string`	The question as it was sent.
`reason`	`string`	`no_usable_answer` — we sent this question and did not end up with an answer we could use; names no actor and makes no claim about the question or the engine. `transcription_failed` — the engine answered and its pages were captured, but the step that reads the companies out of the text failed; the answer is kept as evidence and the slot enters no count until a recompute repairs it. Report either as ‘we could not read this engine’s answer to this question — not counted’. NEVER as ‘not named’ or ‘0 pages’, and never attach ‘failed’ to the question, which the customer wrote. Values: `no_usable_answer`, `transcription_failed`.

item.noAnswerShown[] — each entry

Field	Type	Description
`engine`	`string`	Which engine this question was sent to — we reached it and it showed no answer. Values: `perplexity`, `google_ai_overviews`.
`promptIndex`	`number`	Zero-based position of the question in this check’s question list.
`promptId`	`string`	Stable identifier of the question.
`promptText`	`string`	The question as it was sent.
`reason`	`string`	`no_ai_overview_shown` — Google’s results page for this question carried no AI Overview. A measured fact about the question on Google that day: not a failure, not an answer, and nothing the customer’s question did wrong. Report it as ‘Google showed no AI Overview for this question — not counted’; NEVER as ‘not named’ or a question that failed. This slot is in no count on this response. Values: `no_ai_overview_shown`.

item.engineStatus — object

Field	Type	Description
`perplexity`	`object`	Perplexity. See `item.engineStatus.perplexity`.
`google_ai_overviews`	`object`	Google AI Overviews. See `item.engineStatus.google_ai_overviews`.

item.summary.window — object

Field	Type	Description
`checkIds`	`array<string>`	The published checks pooled into this summary — this check first, then the previous published checks, up to 5 in all. Checks, never days: the interval is a schedule setting, so a three-day and a monthly project with the same window are equally trustworthy. Membership; the count is the array’s length. A single entry is the first check, and every “seen in” figure then reads 1 of 1 — `limits.sentences` says so.
`since`	`string`	When the oldest check in the window started (ISO-8601).
`perEngine`	`object`	Per engine, the checks in the window on which that engine received at least one answer — the M of that engine’s “seen in N of M” on `coreHosts[].seenInChecks`. It moves between checks where an engine showed no answer on a whole check. Keyed by engine. A key is ABSENT for an engine this check did not ask — a check keeps the engine set it was run with — and absent means not measured, never a zero. Never add the engines’ page figures together. See `item.summary.window.perEngine`.

item.summary.perEngine — object

Field	Type	Description
`perplexity`	`object`	Perplexity’s reading of this check. See `item.summary.perEngine.perplexity`.
`google_ai_overviews`	`object`	Google AI Overviews’ reading of this check. See `item.summary.perEngine.google_ai_overviews`.

item.summary.matrix[] — each entry

Field	Type	Description
`promptId`	`string`	Stable identifier of the question, kept across edits of its wording. The handle to join this row with `answers[]`, `coreHosts[].promptIds` and `pages[].promptIds`.
`promptText`	`string`	The question as it was asked — one of the project’s own buying questions.
`intent`	`string`	The question’s angle. The intents are fixed across every project — one question per intent — and the token names the angle and nothing else. Values: `general_choice`, `problem_driven`, `price_constraint`, `company_size`, `feature`, `free_tier`, `safety`, `category_compare`.
`cells`	`object`	One cell per engine this check asked. Keyed by engine. A key is ABSENT for an engine this check did not ask — a check keeps the engine set it was run with — and absent means not measured, never a zero. Never add the engines’ page figures together. See `item.summary.matrix[].cells`.

item.summary.brands[] — each entry

Field	Type	Description
`domain`	`string`	Brand domain — the identity every count is matched on.
`name`	`string`	Brand name as the engines most often wrote it. Decoration: it can change between checks while the domain does not. Never match or join on it.
`isOwn`	`boolean`	Whether this is the customer’s own row. The customer’s row is ALWAYS present, at zero when nothing named them — that zero is a measured finding.
`origin`	`string`	How the brand came to be listed: the customer, a tracked competitor, or a company an engine’s answer named in the window that is not on the competitor list. Values: `customer`, `tracked_competitor`, `named_by_engines`.
`ambiguous`	`boolean`	The name is ordinary language (every word of it is a common word or a word of the category), so every page count for this brand is a range with an open ceiling: quote `floor` and `ceiling` together, never one alone. Inferred, and fails closed.
`rankByPresence`	`number`	Rank by how often the brand is named in answers: one plus the number of brands named more often. Ties SHARE a rank — never break one. Presence, never position: nothing on this surface records how high a brand sat.
`answers`	`object`	Answers in the window that named the brand, pooled across engines as a vote, with the interval. Pooling ANSWERS is legitimate; `perEngine` beside it says which engines back the row, because a brand named by one engine and a brand named by both look identical on the pooled figure. See `item.summary.brands[].answers`.
`answersRetrievingOwnSite`	`object`	Answers in the window in which an engine retrieved at least one page on the brand’s own site, pooled the same way. Being read is not being recommended — compare with `answers`. See `item.summary.brands[].answersRetrievingOwnSite`.
`enginesNaming`	`array<string>`	The engines on which at least one answer in the window named the brand. Membership, never a count.
`perEngine`	`object`	This brand on each engine. A key is absent for an engine that received no answer in the window. Page counts are per engine and are never summed across keys. Keyed by engine. A key is ABSENT for an engine this check did not ask — a check keeps the engine set it was run with — and absent means not measured, never a zero. Never add the engines’ page figures together. See `item.summary.brands[].perEngine`.

item.summary.overlap — object

Field	Type	Description
`hostsInCore`	`number`	Hosts retrieved by at least 2 engines this check — the core, and the ONE cross-engine object on this surface. Render the overlap as ‘only one engine / both / only the other’ with absolute counts; there is deliberately no union total, because no engine produced that list.
`perEngine`	`object`	Per engine this check asked: hosts retrieved, and how many of them no other engine retrieved. Keyed by engine. A key is ABSENT for an engine this check did not ask — a check keeps the engine set it was run with — and absent means not measured, never a zero. Never add the engines’ page figures together. See `item.summary.overlap.perEngine`.

item.summary.funnel — object

Field	Type	Description
`coreHosts`	`number`	Hosts retrieved by at least 2 engines this check — the start of the narrowing, and the universe every other field here is a filter over. Equals `alreadyNamingCustomer + missing + unreadable`, and equals `coreHosts.length`.
`alreadyNamingCustomer`	`number`	Of `coreHosts`: hosts on which at least one page names the customer. Already won — never a target.
`unreadable`	`number`	Of `coreHosts`: hosts on which no page could be read and none names the customer from its title or passage. Listed, never counted as missing — an unread page is never a page the customer is absent from.
`missing`	`number`	Of `coreHosts`: hosts on which at least one page was READ and none names the customer — the pages the customer is genuinely missing from. Equals `missingPublishers + missingCompetitorOwned`. Small on a market leader: ‘already on 19 of the 25 hosts more than one engine read’ is the finding, not an empty state.
`missingPublishers`	`number`	Of `missing`: third-party hosts — publishers, communities, review sites. The work list: addressable by outreach.
`missingCompetitorOwned`	`number`	Of `missing`: a competitor’s own site. Not won by outreach; listed so the narrowing is complete.

item.summary.coreHosts[] — each entry

Field	Type	Description
`host`	`string`	The host, lower-case, without `www.`.
`status`	`string`	The host’s standing for the customer. `already_named` — at least one page on it names the customer: won. `missing` — at least one page was read and none names the customer: the work list, and the only status that supports ‘get onto this’. `unreadable` — no page on the host could be read and none names the customer from its title or passage: listed, not counted, never a target. Values: `already_named`, `missing`, `unreadable`.
`ownership`	`string`	`third_party` — nobody on the leaderboard owns it: a publisher, a community, a review site; addressable. `competitor_owned` — a competitor’s own site; not won by outreach. The customer’s own hosts are never on this list. Values: `third_party`, `competitor_owned`.
`ownedBy`	`string`	The competitor’s domain, present iff `ownership` is `competitor_owned`.
`kind`	`string`	What kind of site the host is, from a dated table of known hosts; anything not in the table is a `publisher`. The kind chooses the action hint; it never changes a count. Values: `review_site`, `community`, `video`, `publisher`.
`kindCheckedAt`	`string`	Present iff the kind came from the table: the date (YYYY-MM-DD) a person last checked what this host is. The kind was true on that date; it is not a claim about today.
`actionHint`	`object`	What the customer can do about this host, as a stable code with its sentence beside it. Render `text` verbatim; it is identical for every host in the same state and names no host’s numbers. `pitch_publisher` — a third-party page we read that does not name the customer. `claim_review_profile` — a review site, whether or not we could read it. `join_community` — a community. `not_addressable_by_text` — a video. `check_manually` — we could not read the page: open it and check. `competitor_owned` — not a target. `already_named` — nothing to do. See `item.summary.coreHosts[].actionHint`.
`engines`	`array<string>`	The engines that retrieved a page on this host this check. Its length is at least 2 — that is what makes the host core. Membership, never a count.
`promptIds`	`array<string>`	The questions whose answers retrieved a page on this host, on any engine. Membership; its length is ‘N of the project’s questions’.
`seenInChecks`	`object`	Per engine, the checks in the window in which this host was in THAT engine’s OWN core — the hosts it retrieved on several of the questions — “seen in N of M”: N is that engine’s array length here, M is `window.perEngine[engine].checkIds.length`. A key is absent for an engine that never had this host in its own core, and absence is never a zero of anything. This is a different core from the row’s: a host is on `coreHosts` for being retrieved by two or more engines at any question count, so a host each engine retrieved once is core by overlap and core to neither alone, and carries an empty object here — correct, and read as “in no engine’s own core”, never as 0 of 1. State N and M only for the engines present; never add the engines together. See `item.summary.coreHosts[].seenInChecks`.
`brandsNamed`	`array<string>`	Domains of the other leaderboard brands named on this host’s pages. Membership; the count is the array’s length.
`pages`	`array<object>`	This host’s pages this check, each with its own read state and naming. A host is `missing` when at least one of these was read and none is `named`. See `item.summary.coreHosts[].pages[]`.

item.summary.ownPageRetrievals[] — each entry

Field	Type	Description
`engine`	`string`	The engine that opened the customer’s site for this answer. Values: `perplexity`, `google_ai_overviews`.
`promptId`	`string`	The question this answer was for. Join on `matrix[].promptId`.
`pages`	`array<object>`	The customer’s own pages the engine retrieved for this answer. See `item.summary.ownPageRetrievals[].pages[]`.
`companiesNamed`	`array<string>`	Domains the answer then named, in order of first mention. The evidence and the recommendation side by side: it opened the customer’s page, then recommended these.
`namedCustomer`	`boolean`	Whether the customer was among them.

item.summary.pages[] — each entry

Field	Type	Description
`key`	`string`	The page’s identity: the canonical form of its address — lower-case host without `www.`, no scheme, tracking parameters removed, no fragment. One page listed under two addresses is read once and counted once.
`url`	`string`	The address as an engine returned it.
`host`	`string`	The host, lower-case, without `www.`.
`title`	`string`	The page’s title — its own when we read it, else the title an engine reported. Absent when neither exists.
`engines`	`array<string>`	The engines that retrieved this page this check. Membership, never a count; its length is how many engines read it.
`promptIds`	`array<string>`	The questions whose answers retrieved this page, on any engine. Join on `matrix[].promptId`.
`ownedBy`	`string`	The domain of the brand whose own site this page is on, when it is on one — the customer’s or a competitor’s. Absent for a third-party page. A page on a brand’s own site is never an independent source naming that brand.
`readState`	`string`	`read` — the body was fetched and read. `unreadable` — we could not read it; `unreadableReason` says what was observed. `not_fetched` — no fetch ever reported for it. Kept apart from `unreadable` because ‘we stopped’ and ‘we could not read it’ are different facts. Values: `read`, `unreadable`, `not_fetched`.
`unreadableReason`	`string`	Present iff `readState` is `unreadable`. Each names what was OBSERVED about the fetch, never a fact about the page’s content: `bot_protection` — the site refused an automated reader; `host_blocked_recently` — not fetched, because the host refused us within the last month; `text_too_short` — a body arrived with too few readable characters; `not_found_page` — the server answered with a page that says the resource does not exist; `consent_wall` — the server answered with a consent interstitial in place of the page; `unreachable` — the host could not be reached; `timeout` — the page did not finish loading in time; `fetch_failed` — any other failure to obtain a body; `unresolved` — the engine returned a redirect we could not resolve to a page. Values: `bot_protection`, `host_blocked_recently`, `text_too_short`, `not_found_page`, `consent_wall`, `unreachable`, `timeout`, `fetch_failed`, `unresolved`.
`fetchedAt`	`string`	When this page’s verdict was taken (ISO-8601) — the age of the evidence. A page is re-read after a month. Absent when not fetched.
`textSource`	`string`	What text this page was judged on. `body` — the fetched page plus the title and passage the engine handed back. `snippet_only` — the title and passage alone, for a page we could not read. `none` — a page with neither. Values: `body`, `snippet_only`, `none`.
`namesCustomer`	`string \| null`	Whether this page’s text names the customer. `named` — it does, and where the name is ordinary language the page corroborates it. `named_uncorroborated` — the name matched but nothing corroborates it; reachable only for an ambiguous brand, and in the ceiling of every range rather than the floor. `not_named` — the page was READ and does not name the customer: the ONLY value that supports ‘get onto this page’. `null` — the page was not read (unreadable or not fetched), so nothing can be said. An unread page can still be `named` from the title and passage the engine handed back: an unread page can prove presence, never absence. Values: `named`, `named_uncorroborated`, `not_named`. Nullable.
`brandsNamed`	`array<string>`	Domains of the OTHER brands on the leaderboard that this page’s text names. Membership; the count is the array’s length. Empty is measured: we looked and found none.
`excerpt`	`string`	A short passage around the first mention of the customer where named, else of the first other brand named — at most 240 characters. Third-party text, quoted as evidence. Absent where no brand was found in the text.

item.summary.limits — object

Field	Type	Description
`pagesRetrieved`	`number`	Distinct pages retrieved this check, across engines. A fact about the FETCH STAGE — one stage over every engine’s pages — and never a headline: it is not a pooled page figure and must not be quoted as ‘the pages the engines read’. Per-engine page counts are under `perEngine`.
`pagesRead`	`number`	Of `pagesRetrieved`: pages whose body we read.
`pagesUnreadable`	`number`	Of `pagesRetrieved`: pages we could not read. `unreadableByReason` breaks it down.
`pagesNotFetched`	`number`	Of `pagesRetrieved`: pages whose fetch never reported.
`pagesTruncated`	`number`	Of `pagesRead`: pages whose text was cut at the length cap before matching.
`unreadableByReason`	`object`	`pagesUnreadable` by reason. A reason with no page is absent. See `item.summary.limits.unreadableByReason`.
`textSourceMix`	`object`	What text each page was judged on, over `pagesRetrieved`. See `item.summary.limits.textSourceMix`.
`sentences`	`array<object>`	The standing sentences under every number here, as payload. Each that applies to this check is included; the app, this API and the MCP tools render them VERBATIM and compose none of their own. Among them: retrieved is not cited; named on a page the engine retrieved, not necessarily in the passage it read; a floor for an ambiguous name; per engine, never pooled; an unread page is listed, not counted; a brand’s own pages are never independent; and, on a first check, that every ‘seen in’ reads 1 of 1. See `item.summary.limits.sentences[]`.

item.answers[].askedIn — object

Field	Type	Description
`locationName`	`string`	The country the question was asked from, as the engine names its markets.
`languageName`	`string`	The language the question was asked in, as the engine names it.

item.answers[].companiesNamed[] — each entry

Field	Type	Description
`rank`	`number`	Position among the companies this answer named — 1 means named first. The ORDER OF FIRST MENTION in `answerText`, computed by CompetLab from the text: the engine assigned no position, so never report it as a rank the engine gave. Stable under filtering.
`name`	`string`	The company’s name as the answer wrote it.
`domain`	`string`	The company’s primary domain, normalised — the identity.

item.answers[].sources[] — each entry

Field	Type	Description
`url`	`string`	Address of the retrieved page, after any redirect the engine returned was resolved.
`domain`	`string`	Domain of the retrieved page.
`title`	`string`	Title of the page as the engine reported it. Empty when the engine reported none.
`snippet`	`string`	The passage the engine handed back for this page beside its answer, verbatim. Present only where the engine reports one; absent means the engine reports none, never that the page has no text.
`publishedAt`	`string`	The page’s publication date as the engine reports it, verbatim — the engine’s string, not a measurement of ours. Present only where the engine reports one.
`resolvedFrom`	`string`	The address the engine actually returned, when `url` is the page it resolved to. Absent on a direct URL.

item.answers[].webSearch — object

Field	Type	Description
`performed`	`boolean`	Whether a live search RAN while the engine wrote this answer. `true` records that a search ran, not that the answer rests on what it found. For an engine read off a results page the search is the request, so it is true whenever the page came back.
`completedSearchCount`	`number`	How many searches the engine reports — each engine’s own quantity, and NOT comparable across engines: one reports billable requests, another only the query strings it lists. Compare an engine against itself over time; never compare engines with each other and never sum them. `0` beside `performed: false`.
`reason`	`string`	Present iff `performed` is false. `NO_SEARCH_ATTEMPTED` — the search tool was offered and went unused. `SEARCH_FAILED` — retrieval was attempted and did not succeed. `NO_SEARCH_REPORTED` — the answer carries no evidence of a search and the engine discloses no cause. None of these means the engine chose to answer from memory; they record that no retrieval is evidenced, and why is not something the data answers. Values: `NO_SEARCH_ATTEMPTED`, `SEARCH_FAILED`, `NO_SEARCH_REPORTED`.

item.engineStatus.perplexity — object

Field	Type	Description
`reported`	`boolean`	Whether this engine stored a result for this check. `false` means nothing came back from it at all. `true` does NOT mean it answered usefully — read the three counts beside it.
`completedAt`	`string \| null`	When this engine finished (ISO-8601). Null when it never reported. Nullable.
`questionsAsked`	`number`	Questions this engine was asked — the check’s own recorded question count.
`answersReceived`	`number`	Of `questionsAsked`: questions it answered — the entries for this engine in `answers`, before any filter. The three counts here are separate facts; state the shortfall against `questionsAsked` as a count, never as a ratio.
`answersAbsent`	`number`	Of `questionsAsked`: questions the engine was read for and showed no answer to — the entries for this engine in `noAnswerShown`. Not failures.
`answersUnmeasured`	`number`	Of `questionsAsked`: questions we could not read the answer to — the entries for this engine in `unansweredQueries`. Our problem, and in no count.

item.engineStatus.google_ai_overviews — object

Field	Type	Description
`reported`	`boolean`	Whether this engine stored a result for this check. `false` means nothing came back from it at all. `true` does NOT mean it answered usefully — read the three counts beside it.
`completedAt`	`string \| null`	When this engine finished (ISO-8601). Null when it never reported. Nullable.
`questionsAsked`	`number`	Questions this engine was asked — the check’s own recorded question count.
`answersReceived`	`number`	Of `questionsAsked`: questions it answered — the entries for this engine in `answers`, before any filter. The three counts here are separate facts; state the shortfall against `questionsAsked` as a count, never as a ratio.
`answersAbsent`	`number`	Of `questionsAsked`: questions the engine was read for and showed no answer to — the entries for this engine in `noAnswerShown`. Not failures.
`answersUnmeasured`	`number`	Of `questionsAsked`: questions we could not read the answer to — the entries for this engine in `unansweredQueries`. Our problem, and in no count.

item.summary.window.perEngine — object

Field	Type	Description
`perplexity`	`object`	Perplexity. Nested — see the Example tab.
`google_ai_overviews`	`object`	Google AI Overviews. Nested — see the Example tab.

item.summary.perEngine.perplexity — object

Field	Type	Description
`engineDataAvailable`	`object`	Present ONLY when this engine produced nothing usable this check. When present, no other field on this object is, and the card for this engine reads as unmeasured with this reason — never as zeros. Nested — see the Example tab.
`answersAsked`	`number`	Questions this engine was asked — the check’s own recorded question count, read off the check.
`answersReceived`	`number`	Questions this engine answered. THE denominator for every ‘answers’ figure on this engine. State the shortfall against `answersAsked` as two facts, never as a ratio.
`answersAbsent`	`number`	Questions the engine was read for and showed no answer to — today, questions Google’s results page carried no AI Overview for. Measured absences: not answers, not failures, and in no denominator.
`answersUnmeasured`	`number`	Questions we could not read this engine’s answer to. In no denominator. Our problem, never a fact about the question or the engine’s view of anyone.
`answersWithoutRetrieval`	`number`	Of `answersReceived`: answers the engine wrote without retrieving any page — it reported no page, and has not said how it answered; say `no pages reported`, never `from memory`. Still answers; they contribute no page.
`answersNamingCustomer`	`number`	Of `answersReceived`: answers that named the customer.
`answersRetrievingOwnPage`	`number`	Of `answersReceived`: answers in which the engine retrieved at least one page on the customer’s own site. Being read is not being recommended — compare with `answersNamingCustomer`.
`pagesRetrieved`	`number`	Distinct pages this engine RETRIEVED this check across its answers. Retrieved, never cited: the engine does not say which it leaned on. This engine’s number alone — never add it to another engine’s.
`hostsRetrieved`	`number`	Distinct hosts among `pagesRetrieved`.
`hostsOnlyThisEngine`	`number`	Of `hostsRetrieved`: hosts no other engine retrieved this check.
`engineCoreSize`	`number`	Hosts this engine retrieved for at least 2 of the questions — this engine’s own core, this check. The universe of that engine’s “seen in N of M” on the core hosts. Small on some engines by nature: quote it beside any “seen in” figure.
`pagesRead`	`number`	Of `pagesRetrieved`: pages whose body we read. The universe of `independentPagesNamingCustomer`.
`pagesUnreadable`	`number`	Of `pagesRetrieved`: pages we could not read. Listed, never counted — an unreadable page is never a page the customer is absent from.
`pagesNotFetched`	`number`	Of `pagesRetrieved`: pages whose fetch never reported. Kept apart from `pagesUnreadable` because ‘we stopped’ and ‘we could not read it’ are different facts.
`ownPagesRetrieved`	`number`	Of `pagesRetrieved`: pages on the customer’s own site, subdomains included. Never counted as independent.
`independentPagesNamingCustomer`	`object`	Independent pages this engine retrieved whose text names the customer — pages not on the customer’s own site — over `pagesRead`. A range: `floor` is the corroborated count, `ceiling` the naive one; equal for an unambiguous name. Named on a page the engine retrieved, not necessarily in the passage it read. Nested — see the Example tab.
`topHosts`	`array<object>`	The hosts this engine leaned on most this check, at most 10, by pages retrieved then host name. Each says which other engines also retrieved it. Nested — see the Example tab.

item.summary.perEngine.google_ai_overviews — object

Field	Type	Description
`engineDataAvailable`	`object`	Present ONLY when this engine produced nothing usable this check. When present, no other field on this object is, and the card for this engine reads as unmeasured with this reason — never as zeros. Nested — see the Example tab.
`answersAsked`	`number`	Questions this engine was asked — the check’s own recorded question count, read off the check.
`answersReceived`	`number`	Questions this engine answered. THE denominator for every ‘answers’ figure on this engine. State the shortfall against `answersAsked` as two facts, never as a ratio.
`answersAbsent`	`number`	Questions the engine was read for and showed no answer to — today, questions Google’s results page carried no AI Overview for. Measured absences: not answers, not failures, and in no denominator.
`answersUnmeasured`	`number`	Questions we could not read this engine’s answer to. In no denominator. Our problem, never a fact about the question or the engine’s view of anyone.
`answersWithoutRetrieval`	`number`	Of `answersReceived`: answers the engine wrote without retrieving any page — it reported no page, and has not said how it answered; say `no pages reported`, never `from memory`. Still answers; they contribute no page.
`answersNamingCustomer`	`number`	Of `answersReceived`: answers that named the customer.
`answersRetrievingOwnPage`	`number`	Of `answersReceived`: answers in which the engine retrieved at least one page on the customer’s own site. Being read is not being recommended — compare with `answersNamingCustomer`.
`pagesRetrieved`	`number`	Distinct pages this engine RETRIEVED this check across its answers. Retrieved, never cited: the engine does not say which it leaned on. This engine’s number alone — never add it to another engine’s.
`hostsRetrieved`	`number`	Distinct hosts among `pagesRetrieved`.
`hostsOnlyThisEngine`	`number`	Of `hostsRetrieved`: hosts no other engine retrieved this check.
`engineCoreSize`	`number`	Hosts this engine retrieved for at least 2 of the questions — this engine’s own core, this check. The universe of that engine’s “seen in N of M” on the core hosts. Small on some engines by nature: quote it beside any “seen in” figure.
`pagesRead`	`number`	Of `pagesRetrieved`: pages whose body we read. The universe of `independentPagesNamingCustomer`.
`pagesUnreadable`	`number`	Of `pagesRetrieved`: pages we could not read. Listed, never counted — an unreadable page is never a page the customer is absent from.
`pagesNotFetched`	`number`	Of `pagesRetrieved`: pages whose fetch never reported. Kept apart from `pagesUnreadable` because ‘we stopped’ and ‘we could not read it’ are different facts.
`ownPagesRetrieved`	`number`	Of `pagesRetrieved`: pages on the customer’s own site, subdomains included. Never counted as independent.
`independentPagesNamingCustomer`	`object`	Independent pages this engine retrieved whose text names the customer — pages not on the customer’s own site — over `pagesRead`. A range: `floor` is the corroborated count, `ceiling` the naive one; equal for an unambiguous name. Named on a page the engine retrieved, not necessarily in the passage it read. Nested — see the Example tab.
`topHosts`	`array<object>`	The hosts this engine leaned on most this check, at most 10, by pages retrieved then host name. Each says which other engines also retrieved it. Nested — see the Example tab.

item.summary.matrix[].cells — object

Field	Type	Description
`perplexity`	`object`	Perplexity’s cell for this question. Nested — see the Example tab.
`google_ai_overviews`	`object`	Google AI Overviews’ cell for this question. Nested — see the Example tab.

item.summary.brands[].answers — object

Field	Type	Description
`answersNaming`	`number`	Answers in this scope that named the brand — counted once per answer, matched on the brand’s domain, never on its wording.
`answersReceived`	`number`	The universe `answersNaming` is drawn from: answers received in this scope. Questions that produced no answer, or that the engine was read for and had nothing to show, are not in it. Quote it whenever you quote the share.
`presence`	`number`	`answersNaming / answersReceived × 100`, whole percent. A share of answers received — never of questions asked, and never a probability. Always read with `presenceLow` / `presenceHigh` beside it.
`presenceLow`	`number`	Lower bound of a 95% interval on `presence`, same percent scale. Two brands whose ranges overlap are NOT ordered, whatever their rank says.
`presenceHigh`	`number`	Upper bound of the same interval. See `presenceLow`.

item.summary.brands[].answersRetrievingOwnSite — object

Field	Type	Description
`answersNaming`	`number`	Answers in this scope that named the brand — counted once per answer, matched on the brand’s domain, never on its wording.
`answersReceived`	`number`	The universe `answersNaming` is drawn from: answers received in this scope. Questions that produced no answer, or that the engine was read for and had nothing to show, are not in it. Quote it whenever you quote the share.
`presence`	`number`	`answersNaming / answersReceived × 100`, whole percent. A share of answers received — never of questions asked, and never a probability. Always read with `presenceLow` / `presenceHigh` beside it.
`presenceLow`	`number`	Lower bound of a 95% interval on `presence`, same percent scale. Two brands whose ranges overlap are NOT ordered, whatever their rank says.
`presenceHigh`	`number`	Upper bound of the same interval. See `presenceLow`.

item.summary.brands[].perEngine — object

Field	Type	Description
`perplexity`	`object`	Perplexity’s reading of this brand. Nested — see the Example tab.
`google_ai_overviews`	`object`	Google AI Overviews’ reading of this brand. Nested — see the Example tab.

item.summary.overlap.perEngine — object

Field	Type	Description
`perplexity`	`object`	Perplexity. Nested — see the Example tab.
`google_ai_overviews`	`object`	Google AI Overviews. Nested — see the Example tab.

item.summary.coreHosts[].actionHint — object

Field	Type	Description
`code`	`string`	Stable identifier for this sentence. Branch on it if you need to; never substitute your own wording for it.
`text`	`string`	The sentence to report, in the words every CompetLab surface uses. Render it VERBATIM — do not paraphrase, shorten, or compose your own from the code. It names no row’s numbers on purpose; the fields beside it carry those.

item.summary.coreHosts[].seenInChecks — object

Field	Type	Description
`perplexity`	`object`	Perplexity. Nested — see the Example tab.
`google_ai_overviews`	`object`	Google AI Overviews. Nested — see the Example tab.

item.summary.coreHosts[].pages[] — each entry

Field	Type	Description
`key`	`string`	The page’s identity: the canonical form of its address — lower-case host without `www.`, no scheme, tracking parameters removed, no fragment. One page listed under two addresses is read once and counted once.
`url`	`string`	The address as an engine returned it.
`host`	`string`	The host, lower-case, without `www.`.
`title`	`string`	The page’s title — its own when we read it, else the title an engine reported. Absent when neither exists.
`engines`	`array<string>`	The engines that retrieved this page this check. Membership, never a count; its length is how many engines read it.
`promptIds`	`array<string>`	The questions whose answers retrieved this page, on any engine. Join on `matrix[].promptId`.
`ownedBy`	`string`	The domain of the brand whose own site this page is on, when it is on one — the customer’s or a competitor’s. Absent for a third-party page. A page on a brand’s own site is never an independent source naming that brand.
`readState`	`string`	`read` — the body was fetched and read. `unreadable` — we could not read it; `unreadableReason` says what was observed. `not_fetched` — no fetch ever reported for it. Kept apart from `unreadable` because ‘we stopped’ and ‘we could not read it’ are different facts. Values: `read`, `unreadable`, `not_fetched`.
`unreadableReason`	`string`	Present iff `readState` is `unreadable`. Each names what was OBSERVED about the fetch, never a fact about the page’s content: `bot_protection` — the site refused an automated reader; `host_blocked_recently` — not fetched, because the host refused us within the last month; `text_too_short` — a body arrived with too few readable characters; `not_found_page` — the server answered with a page that says the resource does not exist; `consent_wall` — the server answered with a consent interstitial in place of the page; `unreachable` — the host could not be reached; `timeout` — the page did not finish loading in time; `fetch_failed` — any other failure to obtain a body; `unresolved` — the engine returned a redirect we could not resolve to a page. Values: `bot_protection`, `host_blocked_recently`, `text_too_short`, `not_found_page`, `consent_wall`, `unreachable`, `timeout`, `fetch_failed`, `unresolved`.
`fetchedAt`	`string`	When this page’s verdict was taken (ISO-8601) — the age of the evidence. A page is re-read after a month. Absent when not fetched.
`textSource`	`string`	What text this page was judged on. `body` — the fetched page plus the title and passage the engine handed back. `snippet_only` — the title and passage alone, for a page we could not read. `none` — a page with neither. Values: `body`, `snippet_only`, `none`.
`namesCustomer`	`string \| null`	Whether this page’s text names the customer. `named` — it does, and where the name is ordinary language the page corroborates it. `named_uncorroborated` — the name matched but nothing corroborates it; reachable only for an ambiguous brand, and in the ceiling of every range rather than the floor. `not_named` — the page was READ and does not name the customer: the ONLY value that supports ‘get onto this page’. `null` — the page was not read (unreadable or not fetched), so nothing can be said. An unread page can still be `named` from the title and passage the engine handed back: an unread page can prove presence, never absence. Values: `named`, `named_uncorroborated`, `not_named`. Nullable.
`brandsNamed`	`array<string>`	Domains of the OTHER brands on the leaderboard that this page’s text names. Membership; the count is the array’s length. Empty is measured: we looked and found none.
`excerpt`	`string`	A short passage around the first mention of the customer where named, else of the first other brand named — at most 240 characters. Third-party text, quoted as evidence. Absent where no brand was found in the text.

item.summary.ownPageRetrievals[].pages[] — each entry

Field	Type	Description
`key`	`string`	The page’s identity — see `pages[].key`.
`url`	`string`	The address as the engine returned it.
`title`	`string`	The page’s title, when known.

item.summary.limits.unreadableByReason — object

Field	Type	Description
`bot_protection`	`number`	Pages a site refused to an automated reader.
`host_blocked_recently`	`number`	Pages recorded without a request, because their host refused us within the last month.
`text_too_short`	`number`	Pages whose body carried too few readable characters.
`not_found_page`	`number`	Pages the server answered with a page saying the resource does not exist. Listed, never counted.
`consent_wall`	`number`	Pages the server answered with a consent interstitial in place of the page. Listed, never counted.
`unreachable`	`number`	Pages whose host could not be reached.
`timeout`	`number`	Pages that did not finish loading in time.
`fetch_failed`	`number`	Pages that failed to fetch for any other reason.
`unresolved`	`number`	Redirect wrappers an engine returned that we could not resolve to a page.

item.summary.limits.textSourceMix — object

Field	Type	Description
`body`	`number`	Pages judged on the fetched body plus the engine’s title and passage.
`snippet_only`	`number`	Pages judged on the engine’s title and passage alone, because the body could not be read.
`none`	`number`	Pages with neither.

item.summary.limits.sentences[] — each entry

Field	Type	Description
`code`	`string`	Stable identifier for this sentence. Branch on it if you need to; never substitute your own wording for it.
`text`	`string`	The sentence to report, in the words every CompetLab surface uses. Render it VERBATIM — do not paraphrase, shorten, or compose your own from the code. It names no row’s numbers on purpose; the fields beside it carry those.

Errors

Status	Code	Meaning
400	`invalid_parameters` · `invalid_run_id` · `invalid_check_id`	Bad request — the payload failed validation.
401	`api_key_missing` · `api_key_invalid` · `api_key_revoked` · `api_key_expired` · `insufficient_scope`	The `CL-API-Key` is missing, malformed, revoked, expired, or lacks the required scope.
404	`project_not_found` · `no_data_available`	No `project` matches the id in the path.

Every error uses the shared { error: { code, message, status } } envelope; code is one of the values listed above.


{
  "error": {
    "code": "invalid_parameters",
    "message": "Domain is required",
    "status": 400
  }
}

Get AI sources check history

GET/v1/projects/{projectId}/ai-sources/history

Get paginated history of AI Sources checks for a project. Each row carries, per engine, the four measured figures for that check — answers received, answers naming the customer, pages read, and independent pages naming the customer — and the funnel from core hosts to hosts the customer is genuinely missing from. Nothing is summed across engines; an engine absent from a row was not asked on that check or produced nothing usable on it, and absent means not measured, never zero. Counts are counts, never rates: report figures as n of N answers and never as a percentage or a share — the question set is small by design, and a share computed from it is false precision. Only published checks are listed: a check abandoned because no engine produced a usable answer never publishes and does not appear here. Uses checkId, not runId — the unit of this dimension is a check, one cycle of every question against every engine. The full summary for a row is on the check-detail route. This page is subject to a response size cap: when truncated is true, whole entries were dropped from the end of items and pagination.hasMore does NOT account for them — lower limit to see the rest rather than paging forward, which would skip them.

Returns · { items, pagination, truncated }

Path parameters

Name	Type	Description
`projectId`	`string`	Project ID

Query parameters

Name	Type	Required	Description
`page`	`number`		Page number (1-indexed). Default `1`. Min `1`.
`limit`	`number`		Number of items per page. Default `20`. Range `1`–`100`.

Request


curl "https://api.competlab.com/v1/projects/507f1f77bcf86cd799439011/ai-sources/history?page=1&limit=20" \
  -H "CL-API-Key: YOUR_COMPETLAB_API_KEY"

Response 200 OK

Example


{
  "items": [
    {
      "checkId": "507f1f77bcf86cd799439012",
      "completedAt": "2026-09-04T10:00:00.000Z",
      "perEngine": {
        "perplexity": {
          "answersReceived": 7,
          "answersNamingCustomer": 3,
          "pagesRead": 118,
          "independentPagesNamingCustomer": {
            "floor": 4,
            "ceiling": 6
          }
        },
        "google_ai_overviews": {
          "answersReceived": 7,
          "answersNamingCustomer": 3,
          "pagesRead": 118,
          "independentPagesNamingCustomer": {
            "floor": 4,
            "ceiling": 6
          }
        }
      },
      "funnel": {
        "coreHosts": 25,
        "alreadyNamingCustomer": 19,
        "unreadable": 2,
        "missing": 4,
        "missingPublishers": 3,
        "missingCompetitorOwned": 1
      }
    }
  ],
  "pagination": {
    "page": 1,
    "limit": 20,
    "total": 47,
    "totalPages": 3,
    "hasMore": true
  },
  "truncated": false
}

Fields

items[] — each entry

Field	Type	Description
`checkId`	`string`	Check ID — pass it to the check-detail route.
`completedAt`	`string`	When this check published (ISO-8601).
`perEngine`	`object`	Per engine, the four measured figures for this check. A key is ABSENT for an engine the check did not ask, and for an engine that produced nothing usable on it — absent means not measured, never a zero. Nothing here is summed across engines. See `items[].perEngine`.
`funnel`	`object`	The narrowing from core hosts to hosts the customer is genuinely missing from, on this check. See `items[].funnel`.

pagination — object

Field	Type	Description
`page`	`number`	Current page number (1-indexed)
`limit`	`number`	Number of items per page
`total`	`number`	Total number of items across all pages
`totalPages`	`number`	Total number of pages
`hasMore`	`boolean`	Whether more pages exist after the current page

items[].perEngine — object

Field	Type	Description
`perplexity`	`object`	Perplexity. See `items[].perEngine.perplexity`.
`google_ai_overviews`	`object`	Google AI Overviews. See `items[].perEngine.google_ai_overviews`.

items[].funnel — object

Field	Type	Description
`coreHosts`	`number`	Hosts retrieved by at least 2 engines this check — the start of the narrowing, and the universe every other field here is a filter over. Equals `alreadyNamingCustomer + missing + unreadable`, and equals `coreHosts.length`.
`alreadyNamingCustomer`	`number`	Of `coreHosts`: hosts on which at least one page names the customer. Already won — never a target.
`unreadable`	`number`	Of `coreHosts`: hosts on which no page could be read and none names the customer from its title or passage. Listed, never counted as missing — an unread page is never a page the customer is absent from.
`missing`	`number`	Of `coreHosts`: hosts on which at least one page was READ and none names the customer — the pages the customer is genuinely missing from. Equals `missingPublishers + missingCompetitorOwned`. Small on a market leader: ‘already on 19 of the 25 hosts more than one engine read’ is the finding, not an empty state.
`missingPublishers`	`number`	Of `missing`: third-party hosts — publishers, communities, review sites. The work list: addressable by outreach.
`missingCompetitorOwned`	`number`	Of `missing`: a competitor’s own site. Not won by outreach; listed so the narrowing is complete.

items[].perEngine.perplexity — object

Field	Type	Description
`answersReceived`	`number`	Questions this engine answered on this check — the denominator of `answersNamingCustomer`.
`answersNamingCustomer`	`number`	Of `answersReceived`: answers that named the customer.
`pagesRead`	`number`	Pages this engine retrieved on this check whose body we read — the universe of `independentPagesNamingCustomer`.
`independentPagesNamingCustomer`	`object`	Independent pages this engine retrieved on this check whose text names the customer, over `pagesRead`. A range; this engine’s number alone. See `items[].perEngine.perplexity.independentPagesNamingCustomer`.

items[].perEngine.google_ai_overviews — object

Field	Type	Description
`answersReceived`	`number`	Questions this engine answered on this check — the denominator of `answersNamingCustomer`.
`answersNamingCustomer`	`number`	Of `answersReceived`: answers that named the customer.
`pagesRead`	`number`	Pages this engine retrieved on this check whose body we read — the universe of `independentPagesNamingCustomer`.
`independentPagesNamingCustomer`	`object`	Independent pages this engine retrieved on this check whose text names the customer, over `pagesRead`. A range; this engine’s number alone. See `items[].perEngine.google_ai_overviews.independentPagesNamingCustomer`.

items[].perEngine.perplexity.independentPagesNamingCustomer — object

Field	Type	Description
`floor`	`number`	The corroborated count. Where a brand’s name is also ordinary language, a page counts here only when it corroborates the name — the brand’s domain in its text or its links, or its name in the page title. For a brand whose name is unambiguous, `floor` equals `ceiling`.
`ceiling`	`number`	The naive count: every page whose text matches the name, corroborated or not. Never quote it alone for an ambiguous brand; quote the pair as a range.

items[].perEngine.google_ai_overviews.independentPagesNamingCustomer — object

Field	Type	Description
`floor`	`number`	The corroborated count. Where a brand’s name is also ordinary language, a page counts here only when it corroborates the name — the brand’s domain in its text or its links, or its name in the page title. For a brand whose name is unambiguous, `floor` equals `ceiling`.
`ceiling`	`number`	The naive count: every page whose text matches the name, corroborated or not. Never quote it alone for an ambiguous brand; quote the pair as a range.

Paginated — pass page and limit query parameters and follow pagination.hasMore to page through the full set.

Errors

Status	Code	Meaning
401	`api_key_missing` · `api_key_invalid` · `api_key_revoked` · `api_key_expired` · `insufficient_scope`	The `CL-API-Key` is missing, malformed, revoked, expired, or lacks the required scope.

Every error uses the shared { error: { code, message, status } } envelope; code is one of the values listed above.


{
  "error": {
    "code": "api_key_invalid",
    "message": "Invalid API key",
    "status": 401
  }
}

Get AI sources data for a specific check

GET/v1/projects/{projectId}/ai-sources/history/{checkId}

Get full AI Sources data for a specific historical check: the summary stored on that check under summary, in exactly the dashboard’s shape and read the same way — per engine, never pooled; retrieved, never cited; only a missing core host is a page to get onto; actionHint.text and summary.limits.sentences are rendered verbatim. The window it pools over (up to 5 published checks, this one first) is on summary.window. Set includeAnswers=true to also get what the engines actually said: every question sent, the answer text, the companies read out of it in order of first mention, and the pages the engine retrieved with the passage it handed back for each — plus engineStatus, one entry per engine the check asked. That block is large; narrow with engine= (one engine) or promptIndex= (one question). No filter changes any number under summary; ranks are stable under filtering. Counts are counts, never rates: report figures as n of N answers and never as a percentage or a share — the question set is small by design, and a share computed from it is false precision. Questions we could not read are listed separately in unansweredQueries, and questions an engine was read for and showed no answer to in noAnswerShown — three different facts, and none of them is ‘not named’. A malformed id answers 400 invalid_check_id; an id that does not belong to this project answers 404 check_not_found; a check that exists but has no published summary — still running, or abandoned (no engine produced a usable answer, or the page stage could not be closed) — answers 404 run_not_summarized, which is a different fact from check_not_found: the check exists, it has nothing to report.

Returns · { item }

Path parameters

Name	Type	Description
`checkId`	`string`	Check ID
`projectId`	`string`	Project ID

Query parameters

Name	Type	Description
`includeAnswers`	`boolean`	Set true to include what the engines actually said — every question sent, the answer text, the companies read out of it in order of first mention, and the pages the engine RETRIEVED to write it, with the passage it handed back for each. Off by default because the block is large: up to 8 answers per engine, each carrying the engine’s full retrieved page list, which for a searching engine can run to dozens of pages with a passage each. Narrow with `engine=` or `promptIndex=` rather than fetching everything. The answer text is the engine’s wording, not CompetLab’s assessment; the pages are retrieved, never cited. Default `false`.
`engine`	`string`	Return only this engine’s answers. Requires `includeAnswers=true`. Narrows `answers`, `unansweredQueries` and `noAnswerShown`; changes nothing under `summary` and does not narrow `engineStatus`. Values: `perplexity`, `google_ai_overviews`.
`promptIndex`	`number`	Return only the answers for this question, across every engine. Requires `includeAnswers=true`. Zero-based: the question’s position in the check’s question list, matching `promptIndex` on each answer.

Request


curl "https://api.competlab.com/v1/projects/507f1f77bcf86cd799439011/ai-sources/history/507f1f77bcf86cd799439012?includeAnswers=false&engine=perplexity&promptIndex=0" \
  -H "CL-API-Key: YOUR_COMPETLAB_API_KEY"

Response 200 OK

Example


{
  "item": {
    "checkId": "507f1f77bcf86cd799439012",
    "completedAt": "2026-09-04T10:00:00.000Z",
    "summary": {
      "customerDomain": "mycompany.com",
      "verdict": "missing_from_most_core_hosts",
      "window": {
        "checkIds": [
          "507f1f77bcf86cd799439012",
          "507f1f77bcf86cd799439011"
        ],
        "since": "2026-08-01T10:00:00.000Z",
        "perEngine": {
          "perplexity": {
            "checkIds": [
              "507f1f77bcf86cd799439012"
            ]
          },
          "google_ai_overviews": {
            "checkIds": [
              "507f1f77bcf86cd799439012"
            ]
          }
        }
      },
      "perEngine": {
        "perplexity": {
          "engineDataAvailable": {
            "available": false,
            "reason": "no_usable_answer"
          },
          "answersAsked": 8,
          "answersReceived": 7,
          "answersAbsent": 1,
          "answersUnmeasured": 0,
          "answersWithoutRetrieval": 1,
          "answersNamingCustomer": 3,
          "answersRetrievingOwnPage": 4,
          "pagesRetrieved": 131,
          "hostsRetrieved": 58,
          "hostsOnlyThisEngine": 33,
          "engineCoreSize": 35,
          "pagesRead": 118,
          "pagesUnreadable": 9,
          "pagesNotFetched": 4,
          "ownPagesRetrieved": 6,
          "independentPagesNamingCustomer": {
            "floor": 4,
            "ceiling": 6
          },
          "topHosts": [
            {
              "host": "g2.com",
              "pages": 3,
              "promptIds": [],
              "alsoRetrievedBy": [
                "google_ai_overviews"
              ]
            }
          ]
        },
        "google_ai_overviews": {
          "engineDataAvailable": {
            "available": false,
            "reason": "no_usable_answer"
          },
          "answersAsked": 8,
          "answersReceived": 7,
          "answersAbsent": 1,
          "answersUnmeasured": 0,
          "answersWithoutRetrieval": 1,
          "answersNamingCustomer": 3,
          "answersRetrievingOwnPage": 4,
          "pagesRetrieved": 131,
          "hostsRetrieved": 58,
          "hostsOnlyThisEngine": 33,
          "engineCoreSize": 35,
          "pagesRead": 118,
          "pagesUnreadable": 9,
          "pagesNotFetched": 4,
          "ownPagesRetrieved": 6,
          "independentPagesNamingCustomer": {
            "floor": 4,
            "ceiling": 6
          },
          "topHosts": [
            {
              "host": "g2.com",
              "pages": 3,
              "promptIds": [],
              "alsoRetrievedBy": [
                "google_ai_overviews"
              ]
            }
          ]
        }
      },
      "matrix": [
        {
          "promptId": "7f3c1a2e-9d44-4c8b-9f0e-2b6a5c9d7e10",
          "promptText": "Which email warm-up tools are safe to use with a Google Workspace domain?",
          "intent": "safety",
          "cells": {
            "perplexity": {
              "state": "answered",
              "reason": "no_ai_overview_shown",
              "named": false,
              "companiesNamed": 6,
              "webSearch": {},
              "pagesRetrieved": 14,
              "pagesRead": 12,
              "pagesUnreadable": 1,
              "pagesNotFetched": 1,
              "pagesNamingCustomer": {},
              "ownPagesRetrieved": 1
            },
            "google_ai_overviews": {
              "state": "answered",
              "reason": "no_ai_overview_shown",
              "named": false,
              "companiesNamed": 6,
              "webSearch": {},
              "pagesRetrieved": 14,
              "pagesRead": 12,
              "pagesUnreadable": 1,
              "pagesNotFetched": 1,
              "pagesNamingCustomer": {},
              "ownPagesRetrieved": 1
            }
          }
        }
      ],
      "brands": [
        {
          "domain": "competitor.com",
          "name": "Competitor Inc",
          "isOwn": false,
          "origin": "tracked_competitor",
          "ambiguous": false,
          "rankByPresence": 2,
          "answers": {
            "answersNaming": 5,
            "answersReceived": 16,
            "presence": 31,
            "presenceLow": 14,
            "presenceHigh": 56
          },
          "answersRetrievingOwnSite": {
            "answersNaming": 5,
            "answersReceived": 16,
            "presence": 31,
            "presenceLow": 14,
            "presenceHigh": 56
          },
          "enginesNaming": [
            "perplexity"
          ],
          "perEngine": {
            "perplexity": {
              "answers": {},
              "independentPagesNaming": {},
              "pagesRead": 118,
              "ownPagesRetrieved": 2
            },
            "google_ai_overviews": {
              "answers": {},
              "independentPagesNaming": {},
              "pagesRead": 118,
              "ownPagesRetrieved": 2
            }
          }
        }
      ],
      "overlap": {
        "hostsInCore": 25,
        "perEngine": {
          "perplexity": {
            "hostsRetrieved": 58,
            "hostsOnlyThisEngine": 33
          },
          "google_ai_overviews": {
            "hostsRetrieved": 58,
            "hostsOnlyThisEngine": 33
          }
        }
      },
      "funnel": {
        "coreHosts": 25,
        "alreadyNamingCustomer": 19,
        "unreadable": 2,
        "missing": 4,
        "missingPublishers": 3,
        "missingCompetitorOwned": 1
      },
      "coreHosts": [
        {
          "host": "g2.com",
          "status": "missing",
          "ownership": "third_party",
          "ownedBy": "competitor.com",
          "kind": "review_site",
          "kindCheckedAt": "2026-09-04",
          "actionHint": {
            "code": "pitch_publisher",
            "text": "Pitch an update to this page: it names your competitors and not you."
          },
          "engines": [
            "perplexity",
            "google_ai_overviews"
          ],
          "promptIds": [
            "value"
          ],
          "seenInChecks": {
            "perplexity": {
              "checkIds": [
                "507f1f77bcf86cd799439012"
              ]
            },
            "google_ai_overviews": {
              "checkIds": [
                "507f1f77bcf86cd799439012"
              ]
            }
          },
          "brandsNamed": [
            "competitor.com"
          ],
          "pages": [
            {
              "key": "g2.com/categories/email-warm-up",
              "url": "https://www.g2.com/categories/email-warm-up",
              "host": "g2.com",
              "title": "Best Email Warm-Up Software",
              "engines": [
                "perplexity",
                "google_ai_overviews"
              ],
              "promptIds": [],
              "ownedBy": "competitor.com",
              "readState": "read",
              "unreadableReason": "bot_protection",
              "fetchedAt": "2026-09-04T10:00:00.000Z",
              "textSource": "body",
              "namesCustomer": "not_named",
              "brandsNamed": [
                "competitor.com",
                "rival.io"
              ],
              "excerpt": "…the most reliable options this year are Competitor Inc and Rival…"
            }
          ]
        }
      ],
      "ownPageRetrievals": [
        {
          "engine": "perplexity",
          "promptId": "7f3c1a2e-9d44-4c8b-9f0e-2b6a5c9d7e10",
          "pages": [
            {
              "key": "mycompany.com/pricing",
              "url": "https://mycompany.com/pricing",
              "title": "Pricing — My Company"
            }
          ],
          "companiesNamed": [
            "competitor.com",
            "rival.io"
          ],
          "namedCustomer": false
        }
      ],
      "pages": [
        {
          "key": "g2.com/categories/email-warm-up",
          "url": "https://www.g2.com/categories/email-warm-up",
          "host": "g2.com",
          "title": "Best Email Warm-Up Software",
          "engines": [
            "perplexity",
            "google_ai_overviews"
          ],
          "promptIds": [
            "value"
          ],
          "ownedBy": "competitor.com",
          "readState": "read",
          "unreadableReason": "bot_protection",
          "fetchedAt": "2026-09-04T10:00:00.000Z",
          "textSource": "body",
          "namesCustomer": "not_named",
          "brandsNamed": [
            "competitor.com",
            "rival.io"
          ],
          "excerpt": "…the most reliable options this year are Competitor Inc and Rival…"
        }
      ],
      "limits": {
        "pagesRetrieved": 153,
        "pagesRead": 140,
        "pagesUnreadable": 9,
        "pagesNotFetched": 4,
        "pagesTruncated": 3,
        "unreadableByReason": {
          "bot_protection": 5,
          "host_blocked_recently": 2,
          "text_too_short": 1,
          "not_found_page": 0,
          "consent_wall": 0,
          "unreachable": 0,
          "timeout": 1,
          "fetch_failed": 0,
          "unresolved": 0
        },
        "textSourceMix": {
          "body": 140,
          "snippet_only": 9,
          "none": 4
        },
        "sentences": [
          {
            "code": "sources_retrieved_not_cited",
            "text": "Retrieved, never cited: these are the pages the engine pulled back while answering; which of them it leaned on is not something the engines disclose."
          }
        ]
      }
    },
    "answers": [
      {
        "engine": "perplexity",
        "promptIndex": 0,
        "promptId": "7f3c1a2e-9d44-4c8b-9f0e-2b6a5c9d7e10",
        "promptText": "Which email warm-up tools are safe to use with a Google Workspace domain?",
        "intent": "safety",
        "askedIn": {
          "locationName": "United States",
          "languageName": "English"
        },
        "answerText": "For a Google Workspace domain, the safest warm-up tools are Competitor Inc and Rival, both of which…",
        "companiesNamed": [
          {
            "rank": 1,
            "name": "Competitor Inc",
            "domain": "competitor.com"
          }
        ],
        "sources": [
          {
            "url": "https://www.g2.com/categories/email-warm-up",
            "domain": "g2.com",
            "title": "Best Email Warm-Up Software",
            "snippet": "Warm-up tools gradually increase sending volume so that mailbox providers…",
            "publishedAt": "2026-05-12",
            "resolvedFrom": "https://www.google.com/goto?url=…"
          }
        ],
        "searchQueries": [
          "email warm-up tools Google Workspace safe"
        ],
        "webSearch": {
          "performed": true,
          "completedSearchCount": 3,
          "reason": "NO_SEARCH_ATTEMPTED"
        }
      }
    ],
    "unansweredQueries": [
      {
        "engine": "perplexity",
        "promptIndex": 5,
        "promptId": "7f3c1a2e-9d44-4c8b-9f0e-2b6a5c9d7e10",
        "promptText": "Which email warm-up tools have a free plan?",
        "reason": "no_usable_answer"
      }
    ],
    "noAnswerShown": [
      {
        "engine": "google_ai_overviews",
        "promptIndex": 3,
        "promptId": "7f3c1a2e-9d44-4c8b-9f0e-2b6a5c9d7e10",
        "promptText": "Which email warm-up tools have a free plan?",
        "reason": "no_ai_overview_shown"
      }
    ],
    "engineStatus": {
      "perplexity": {
        "reported": true,
        "completedAt": "2026-09-04T10:00:00.000Z",
        "questionsAsked": 8,
        "answersReceived": 7,
        "answersAbsent": 1,
        "answersUnmeasured": 0
      },
      "google_ai_overviews": {
        "reported": true,
        "completedAt": "2026-09-04T10:00:00.000Z",
        "questionsAsked": 8,
        "answersReceived": 7,
        "answersAbsent": 1,
        "answersUnmeasured": 0
      }
    },
    "answersTruncated": false
  }
}

Fields

item — object

Field	Type	Description
`checkId`	`string`	Check ID
`completedAt`	`string`	When this check published (ISO-8601).
`summary`	`object`	The summary stored on this check, exactly as the app shows it — the same shape as the dashboard’s `summary`, as of this check. See `item.summary`.
`answers`	`array<object>`	The engines’ answers for this check — one entry per question that came back with an answer, with the pages the engine RETRIEVED to write it. Present ONLY when the request set `includeAnswers=true`; absent otherwise, never an empty array standing in for ‘not requested’. Narrowed by `engine` and `promptIndex` when those are set — they narrow `unansweredQueries` and `noAnswerShown` the same way. No filter changes anything under `summary`: every number there is stored, computed over the whole check, and never recomputed for a filtered view. The pages on an answer are the pages the engine retrieved while answering, never a list of citations — the engines do not disclose which pages they leaned on. Never count pages across engines. ATTRIBUTION: the answer text and the companies read out of it are what that engine said, including about third parties CompetLab does not monitor. Attribute it to the named `engine`; do not republish it as fact. See `item.answers[]`.
`unansweredQueries`	`array<object>`	Questions this check sent that we could not read an answer to. Present ONLY when `includeAnswers=true`. Carried in their own array rather than mixed into `answers`, so that `answers.length` always means answers received. These slots are in no count on this response. Never quote this array’s length as a fraction of anything: state it as a separate fact. See `item.unansweredQueries[]`.
`noAnswerShown`	`array<object>`	Questions an engine was read for and showed no answer to — today, questions Google’s results page carried no AI Overview for. Not a failure and not an answer: in no count on this response. Separate from `unansweredQueries`, which are questions we could not read. Present ONLY when `includeAnswers=true`. Narrowed by `engine` and `promptIndex`. See `item.noAnswerShown[]`.
`engineStatus`	`object`	Per-engine reporting status for this check: whether each engine this check asked stored a result, when, and how its questions split between answered, no answer shown and unmeasured. One entry per engine the check ASKED — an engine is absent when the check did not ask it; absent means not measured, never ‘did not report’. Present ONLY when `includeAnswers=true`. NOT narrowed by the filters: it always describes the whole check. See `item.engineStatus`.
`answersTruncated`	`boolean`	True when the answers payload hit the response size cap and answers were dropped from the end of `answers`. The cut lands on a WHOLE QUESTION ROUND, never inside one: `answers` is ordered question by question across the engines, so truncation drops the highest `promptIndex` values and no engine is ever removed by truncation from a question that is still present — group the array by `promptIndex` and no round is short. Neither `unansweredQueries` nor `noAnswerShown` is ever truncated, so a question absent from `answers` is not a question no engine answered. An engine missing from a question that IS present is in one of those two arrays, not cut; an engine missing from EVERY question stored nothing for this check, which `engineStatus` reports as its three counts summing below `questionsAsked`. An answer’s page list is never partially truncated. `summary` and `engineStatus` are never affected. Narrow with `engine` or `promptIndex` to get a complete view. Present ONLY when `includeAnswers=true`.

item.summary — object

Field	Type	Description
`customerDomain`	`string`	The customer’s domain, normalised.
`verdict`	`string`	The one reading of the customer’s standing this check, decided when the summary was built and carried here as stored — a consumer never re-derives it from the numbers. Each value names a CONDITION over fields on this same object, never a rating: `recommended_nowhere` — no answer in the window named the customer (the row on `brands` with `isOwn: true` has `answers.answersNaming` of 0). `named_on_most_core_hosts` — the customer is named on at least half of the core hosts (`funnel.alreadyNamingCustomer × 2` is at least `funnel.coreHosts`, with a core to speak of): a short work list under this value is the finding, because there is little left to win. `missing_from_most_core_hosts` — otherwise: named somewhere in the window, and absent from most of the hosts more than one engine read. State the condition beside the two counts it rests on; it ranks nobody and is not a score. Values: `recommended_nowhere`, `named_on_most_core_hosts`, `missing_from_most_core_hosts`.
`window`	`object`	The checks this summary pools over, and per engine the checks that engine answered on. See `item.summary.window`.
`perEngine`	`object`	One reading per engine this check asked — Perplexity and Google AI Overviews today. Every number under a key is that engine’s alone. Keyed by engine. A key is ABSENT for an engine this check did not ask — a check keeps the engine set it was run with — and absent means not measured, never a zero. Never add the engines’ page figures together. Nothing on this surface is a citation count, and no field sums the engines’ pages. See `item.summary.perEngine`.
`matrix`	`array<object>`	One row per question the check asked, in question order, with one cell per engine. Two independent facts sit in an answered cell — whether the answer named the customer, and how many retrieved pages name them — and neither is derived from the other. See `item.summary.matrix[]`.
`brands`	`array<object>`	Who else this market names: the customer, the tracked competitors, and every company an engine named in the window. In the ORDER TO RENDER — answers naming desc, then domain — and never re-sorted by anything positional. The customer’s row is always present; read `isOwn` to find it. Ties share a `rankByPresence`. See `item.summary.brands[]`.
`overlap`	`object`	How the engines’ host sets overlap this check — the one cross-engine object. See `item.summary.overlap`.
`funnel`	`object`	The narrowing from ‘hosts more than one engine read’ to ‘hosts the customer is genuinely missing from’, in four counts that are each a filter over `coreHosts`. Show the narrowing, not only its end: on a market leader the end is small because most of the core already names them. See `item.summary.funnel`.
`coreHosts`	`array<object>`	Every core host with its standing — the work list (`missing`), the won list (`already_named`) and the hosts we could not read (`unreadable`) in ONE array, in the ORDER TO RENDER: the work list first (third-party before competitor-owned), then unreadable, then already named. `status` is the column that separates them. Never tell a customer to get onto a host whose status is not `missing`. See `item.summary.coreHosts[]`.
`ownPageRetrievals`	`array<object>`	Every answer in which an engine opened one of the customer’s own pages, with who it then recommended. Empty is measured: no engine retrieved a page of theirs this check. See `item.summary.ownPageRetrievals[]`.
`pages`	`array<object>`	Every page retrieved this check by any engine, with its read state and naming — the pages-read list. Each says which engines retrieved it; the list itself is a fetch-stage inventory, not a per-engine count, so never quote its length as an engine’s number. See `item.summary.pages[]`.
`limits`	`object`	What the numbers rest on, and the sentences to render beside them. See `item.summary.limits`.

item.answers[] — each entry

Field	Type	Description
`engine`	`string`	Which engine produced this answer. Values: `perplexity`, `google_ai_overviews`.
`promptIndex`	`number`	Zero-based position of the question in this check’s question list, as the engine’s answers are stored. The value the `promptIndex` filter matches. Stable under filtering. Use `promptId` to join this row with the summary’s matrix and pages.
`promptId`	`string`	Stable identifier of the question, kept across edits of its wording.
`promptText`	`string`	The question as it was sent — one of the project’s own buying questions.
`intent`	`string`	The question’s angle. Fixed across every project; the token names the angle and nothing else. Values: `general_choice`, `problem_driven`, `price_constraint`, `company_size`, `feature`, `free_tier`, `safety`, `category_compare`.
`askedIn`	`object`	Where the engine was asked. Both engines are asked from the project’s market, so the retrieved pages are that market’s web. See `item.answers[].askedIn`.
`answerText`	`string`	The answer as the engine wrote it, verbatim — the evidence the companies below were read from. Engine prose, attributed to the engine that wrote it, never to CompetLab.
`companiesNamed`	`array<object>`	The companies the answer named, in order of first mention in `answerText`, each verified present in the text. An EMPTY list is a measurement, never missing data: the engine answered and named no company — it explained a concept and recommended nobody. A question that produced no answer is not here at all; it is in `unansweredQueries`, and a question the engine showed no answer to is in `noAnswerShown`. See `item.answers[].companiesNamed[]`.
`sources`	`array<object>`	The pages the engine RETRIEVED while it answered, distinct by URL, each with whatever the engine reports about it. Retrieved, never cited: the engine does not say which of them it leaned on, and nothing here may be called a citation. ABSENT means the engine reported no retrieval for this answer — it reported none, and has not said how it answered — never that it read nothing we could see. An EMPTY array is the measured ‘it reported retrieving nothing’. See `item.answers[].sources[]`.
`searchQueries`	`array<string>`	The search strings the engine itself issued, as it reports them. Absent where the engine reports none. Strings, not a search count: one search can fan out into several.
`webSearch`	`object`	Whether a live search ran for this answer. Absent means not recorded, never ‘did not search’. `performed: false` is the engine’s own record that no search ran, kept as it is. See `item.answers[].webSearch`.

item.unansweredQueries[] — each entry

Field	Type	Description
`engine`	`string`	Which engine this question was sent to. Values: `perplexity`, `google_ai_overviews`.
`promptIndex`	`number`	Zero-based position of the question in this check’s question list.
`promptId`	`string`	Stable identifier of the question.
`promptText`	`string`	The question as it was sent.
`reason`	`string`	`no_usable_answer` — we sent this question and did not end up with an answer we could use; names no actor and makes no claim about the question or the engine. `transcription_failed` — the engine answered and its pages were captured, but the step that reads the companies out of the text failed; the answer is kept as evidence and the slot enters no count until a recompute repairs it. Report either as ‘we could not read this engine’s answer to this question — not counted’. NEVER as ‘not named’ or ‘0 pages’, and never attach ‘failed’ to the question, which the customer wrote. Values: `no_usable_answer`, `transcription_failed`.

item.noAnswerShown[] — each entry

Field	Type	Description
`engine`	`string`	Which engine this question was sent to — we reached it and it showed no answer. Values: `perplexity`, `google_ai_overviews`.
`promptIndex`	`number`	Zero-based position of the question in this check’s question list.
`promptId`	`string`	Stable identifier of the question.
`promptText`	`string`	The question as it was sent.
`reason`	`string`	`no_ai_overview_shown` — Google’s results page for this question carried no AI Overview. A measured fact about the question on Google that day: not a failure, not an answer, and nothing the customer’s question did wrong. Report it as ‘Google showed no AI Overview for this question — not counted’; NEVER as ‘not named’ or a question that failed. This slot is in no count on this response. Values: `no_ai_overview_shown`.

item.engineStatus — object

Field	Type	Description
`perplexity`	`object`	Perplexity. See `item.engineStatus.perplexity`.
`google_ai_overviews`	`object`	Google AI Overviews. See `item.engineStatus.google_ai_overviews`.

item.summary.window — object

Field	Type	Description
`checkIds`	`array<string>`	The published checks pooled into this summary — this check first, then the previous published checks, up to 5 in all. Checks, never days: the interval is a schedule setting, so a three-day and a monthly project with the same window are equally trustworthy. Membership; the count is the array’s length. A single entry is the first check, and every “seen in” figure then reads 1 of 1 — `limits.sentences` says so.
`since`	`string`	When the oldest check in the window started (ISO-8601).
`perEngine`	`object`	Per engine, the checks in the window on which that engine received at least one answer — the M of that engine’s “seen in N of M” on `coreHosts[].seenInChecks`. It moves between checks where an engine showed no answer on a whole check. Keyed by engine. A key is ABSENT for an engine this check did not ask — a check keeps the engine set it was run with — and absent means not measured, never a zero. Never add the engines’ page figures together. See `item.summary.window.perEngine`.

item.summary.perEngine — object

Field	Type	Description
`perplexity`	`object`	Perplexity’s reading of this check. See `item.summary.perEngine.perplexity`.
`google_ai_overviews`	`object`	Google AI Overviews’ reading of this check. See `item.summary.perEngine.google_ai_overviews`.

item.summary.matrix[] — each entry

Field	Type	Description
`promptId`	`string`	Stable identifier of the question, kept across edits of its wording. The handle to join this row with `answers[]`, `coreHosts[].promptIds` and `pages[].promptIds`.
`promptText`	`string`	The question as it was asked — one of the project’s own buying questions.
`intent`	`string`	The question’s angle. The intents are fixed across every project — one question per intent — and the token names the angle and nothing else. Values: `general_choice`, `problem_driven`, `price_constraint`, `company_size`, `feature`, `free_tier`, `safety`, `category_compare`.
`cells`	`object`	One cell per engine this check asked. Keyed by engine. A key is ABSENT for an engine this check did not ask — a check keeps the engine set it was run with — and absent means not measured, never a zero. Never add the engines’ page figures together. See `item.summary.matrix[].cells`.

item.summary.brands[] — each entry

Field	Type	Description
`domain`	`string`	Brand domain — the identity every count is matched on.
`name`	`string`	Brand name as the engines most often wrote it. Decoration: it can change between checks while the domain does not. Never match or join on it.
`isOwn`	`boolean`	Whether this is the customer’s own row. The customer’s row is ALWAYS present, at zero when nothing named them — that zero is a measured finding.
`origin`	`string`	How the brand came to be listed: the customer, a tracked competitor, or a company an engine’s answer named in the window that is not on the competitor list. Values: `customer`, `tracked_competitor`, `named_by_engines`.
`ambiguous`	`boolean`	The name is ordinary language (every word of it is a common word or a word of the category), so every page count for this brand is a range with an open ceiling: quote `floor` and `ceiling` together, never one alone. Inferred, and fails closed.
`rankByPresence`	`number`	Rank by how often the brand is named in answers: one plus the number of brands named more often. Ties SHARE a rank — never break one. Presence, never position: nothing on this surface records how high a brand sat.
`answers`	`object`	Answers in the window that named the brand, pooled across engines as a vote, with the interval. Pooling ANSWERS is legitimate; `perEngine` beside it says which engines back the row, because a brand named by one engine and a brand named by both look identical on the pooled figure. See `item.summary.brands[].answers`.
`answersRetrievingOwnSite`	`object`	Answers in the window in which an engine retrieved at least one page on the brand’s own site, pooled the same way. Being read is not being recommended — compare with `answers`. See `item.summary.brands[].answersRetrievingOwnSite`.
`enginesNaming`	`array<string>`	The engines on which at least one answer in the window named the brand. Membership, never a count.
`perEngine`	`object`	This brand on each engine. A key is absent for an engine that received no answer in the window. Page counts are per engine and are never summed across keys. Keyed by engine. A key is ABSENT for an engine this check did not ask — a check keeps the engine set it was run with — and absent means not measured, never a zero. Never add the engines’ page figures together. See `item.summary.brands[].perEngine`.

item.summary.overlap — object

Field	Type	Description
`hostsInCore`	`number`	Hosts retrieved by at least 2 engines this check — the core, and the ONE cross-engine object on this surface. Render the overlap as ‘only one engine / both / only the other’ with absolute counts; there is deliberately no union total, because no engine produced that list.
`perEngine`	`object`	Per engine this check asked: hosts retrieved, and how many of them no other engine retrieved. Keyed by engine. A key is ABSENT for an engine this check did not ask — a check keeps the engine set it was run with — and absent means not measured, never a zero. Never add the engines’ page figures together. See `item.summary.overlap.perEngine`.

item.summary.funnel — object

Field	Type	Description
`coreHosts`	`number`	Hosts retrieved by at least 2 engines this check — the start of the narrowing, and the universe every other field here is a filter over. Equals `alreadyNamingCustomer + missing + unreadable`, and equals `coreHosts.length`.
`alreadyNamingCustomer`	`number`	Of `coreHosts`: hosts on which at least one page names the customer. Already won — never a target.
`unreadable`	`number`	Of `coreHosts`: hosts on which no page could be read and none names the customer from its title or passage. Listed, never counted as missing — an unread page is never a page the customer is absent from.
`missing`	`number`	Of `coreHosts`: hosts on which at least one page was READ and none names the customer — the pages the customer is genuinely missing from. Equals `missingPublishers + missingCompetitorOwned`. Small on a market leader: ‘already on 19 of the 25 hosts more than one engine read’ is the finding, not an empty state.
`missingPublishers`	`number`	Of `missing`: third-party hosts — publishers, communities, review sites. The work list: addressable by outreach.
`missingCompetitorOwned`	`number`	Of `missing`: a competitor’s own site. Not won by outreach; listed so the narrowing is complete.

item.summary.coreHosts[] — each entry

Field	Type	Description
`host`	`string`	The host, lower-case, without `www.`.
`status`	`string`	The host’s standing for the customer. `already_named` — at least one page on it names the customer: won. `missing` — at least one page was read and none names the customer: the work list, and the only status that supports ‘get onto this’. `unreadable` — no page on the host could be read and none names the customer from its title or passage: listed, not counted, never a target. Values: `already_named`, `missing`, `unreadable`.
`ownership`	`string`	`third_party` — nobody on the leaderboard owns it: a publisher, a community, a review site; addressable. `competitor_owned` — a competitor’s own site; not won by outreach. The customer’s own hosts are never on this list. Values: `third_party`, `competitor_owned`.
`ownedBy`	`string`	The competitor’s domain, present iff `ownership` is `competitor_owned`.
`kind`	`string`	What kind of site the host is, from a dated table of known hosts; anything not in the table is a `publisher`. The kind chooses the action hint; it never changes a count. Values: `review_site`, `community`, `video`, `publisher`.
`kindCheckedAt`	`string`	Present iff the kind came from the table: the date (YYYY-MM-DD) a person last checked what this host is. The kind was true on that date; it is not a claim about today.
`actionHint`	`object`	What the customer can do about this host, as a stable code with its sentence beside it. Render `text` verbatim; it is identical for every host in the same state and names no host’s numbers. `pitch_publisher` — a third-party page we read that does not name the customer. `claim_review_profile` — a review site, whether or not we could read it. `join_community` — a community. `not_addressable_by_text` — a video. `check_manually` — we could not read the page: open it and check. `competitor_owned` — not a target. `already_named` — nothing to do. See `item.summary.coreHosts[].actionHint`.
`engines`	`array<string>`	The engines that retrieved a page on this host this check. Its length is at least 2 — that is what makes the host core. Membership, never a count.
`promptIds`	`array<string>`	The questions whose answers retrieved a page on this host, on any engine. Membership; its length is ‘N of the project’s questions’.
`seenInChecks`	`object`	Per engine, the checks in the window in which this host was in THAT engine’s OWN core — the hosts it retrieved on several of the questions — “seen in N of M”: N is that engine’s array length here, M is `window.perEngine[engine].checkIds.length`. A key is absent for an engine that never had this host in its own core, and absence is never a zero of anything. This is a different core from the row’s: a host is on `coreHosts` for being retrieved by two or more engines at any question count, so a host each engine retrieved once is core by overlap and core to neither alone, and carries an empty object here — correct, and read as “in no engine’s own core”, never as 0 of 1. State N and M only for the engines present; never add the engines together. See `item.summary.coreHosts[].seenInChecks`.
`brandsNamed`	`array<string>`	Domains of the other leaderboard brands named on this host’s pages. Membership; the count is the array’s length.
`pages`	`array<object>`	This host’s pages this check, each with its own read state and naming. A host is `missing` when at least one of these was read and none is `named`. See `item.summary.coreHosts[].pages[]`.

item.summary.ownPageRetrievals[] — each entry

Field	Type	Description
`engine`	`string`	The engine that opened the customer’s site for this answer. Values: `perplexity`, `google_ai_overviews`.
`promptId`	`string`	The question this answer was for. Join on `matrix[].promptId`.
`pages`	`array<object>`	The customer’s own pages the engine retrieved for this answer. See `item.summary.ownPageRetrievals[].pages[]`.
`companiesNamed`	`array<string>`	Domains the answer then named, in order of first mention. The evidence and the recommendation side by side: it opened the customer’s page, then recommended these.
`namedCustomer`	`boolean`	Whether the customer was among them.

item.summary.pages[] — each entry

Field	Type	Description
`key`	`string`	The page’s identity: the canonical form of its address — lower-case host without `www.`, no scheme, tracking parameters removed, no fragment. One page listed under two addresses is read once and counted once.
`url`	`string`	The address as an engine returned it.
`host`	`string`	The host, lower-case, without `www.`.
`title`	`string`	The page’s title — its own when we read it, else the title an engine reported. Absent when neither exists.
`engines`	`array<string>`	The engines that retrieved this page this check. Membership, never a count; its length is how many engines read it.
`promptIds`	`array<string>`	The questions whose answers retrieved this page, on any engine. Join on `matrix[].promptId`.
`ownedBy`	`string`	The domain of the brand whose own site this page is on, when it is on one — the customer’s or a competitor’s. Absent for a third-party page. A page on a brand’s own site is never an independent source naming that brand.
`readState`	`string`	`read` — the body was fetched and read. `unreadable` — we could not read it; `unreadableReason` says what was observed. `not_fetched` — no fetch ever reported for it. Kept apart from `unreadable` because ‘we stopped’ and ‘we could not read it’ are different facts. Values: `read`, `unreadable`, `not_fetched`.
`unreadableReason`	`string`	Present iff `readState` is `unreadable`. Each names what was OBSERVED about the fetch, never a fact about the page’s content: `bot_protection` — the site refused an automated reader; `host_blocked_recently` — not fetched, because the host refused us within the last month; `text_too_short` — a body arrived with too few readable characters; `not_found_page` — the server answered with a page that says the resource does not exist; `consent_wall` — the server answered with a consent interstitial in place of the page; `unreachable` — the host could not be reached; `timeout` — the page did not finish loading in time; `fetch_failed` — any other failure to obtain a body; `unresolved` — the engine returned a redirect we could not resolve to a page. Values: `bot_protection`, `host_blocked_recently`, `text_too_short`, `not_found_page`, `consent_wall`, `unreachable`, `timeout`, `fetch_failed`, `unresolved`.
`fetchedAt`	`string`	When this page’s verdict was taken (ISO-8601) — the age of the evidence. A page is re-read after a month. Absent when not fetched.
`textSource`	`string`	What text this page was judged on. `body` — the fetched page plus the title and passage the engine handed back. `snippet_only` — the title and passage alone, for a page we could not read. `none` — a page with neither. Values: `body`, `snippet_only`, `none`.
`namesCustomer`	`string \| null`	Whether this page’s text names the customer. `named` — it does, and where the name is ordinary language the page corroborates it. `named_uncorroborated` — the name matched but nothing corroborates it; reachable only for an ambiguous brand, and in the ceiling of every range rather than the floor. `not_named` — the page was READ and does not name the customer: the ONLY value that supports ‘get onto this page’. `null` — the page was not read (unreadable or not fetched), so nothing can be said. An unread page can still be `named` from the title and passage the engine handed back: an unread page can prove presence, never absence. Values: `named`, `named_uncorroborated`, `not_named`. Nullable.
`brandsNamed`	`array<string>`	Domains of the OTHER brands on the leaderboard that this page’s text names. Membership; the count is the array’s length. Empty is measured: we looked and found none.
`excerpt`	`string`	A short passage around the first mention of the customer where named, else of the first other brand named — at most 240 characters. Third-party text, quoted as evidence. Absent where no brand was found in the text.

item.summary.limits — object

Field	Type	Description
`pagesRetrieved`	`number`	Distinct pages retrieved this check, across engines. A fact about the FETCH STAGE — one stage over every engine’s pages — and never a headline: it is not a pooled page figure and must not be quoted as ‘the pages the engines read’. Per-engine page counts are under `perEngine`.
`pagesRead`	`number`	Of `pagesRetrieved`: pages whose body we read.
`pagesUnreadable`	`number`	Of `pagesRetrieved`: pages we could not read. `unreadableByReason` breaks it down.
`pagesNotFetched`	`number`	Of `pagesRetrieved`: pages whose fetch never reported.
`pagesTruncated`	`number`	Of `pagesRead`: pages whose text was cut at the length cap before matching.
`unreadableByReason`	`object`	`pagesUnreadable` by reason. A reason with no page is absent. See `item.summary.limits.unreadableByReason`.
`textSourceMix`	`object`	What text each page was judged on, over `pagesRetrieved`. See `item.summary.limits.textSourceMix`.
`sentences`	`array<object>`	The standing sentences under every number here, as payload. Each that applies to this check is included; the app, this API and the MCP tools render them VERBATIM and compose none of their own. Among them: retrieved is not cited; named on a page the engine retrieved, not necessarily in the passage it read; a floor for an ambiguous name; per engine, never pooled; an unread page is listed, not counted; a brand’s own pages are never independent; and, on a first check, that every ‘seen in’ reads 1 of 1. See `item.summary.limits.sentences[]`.

item.answers[].askedIn — object

Field	Type	Description
`locationName`	`string`	The country the question was asked from, as the engine names its markets.
`languageName`	`string`	The language the question was asked in, as the engine names it.

item.answers[].companiesNamed[] — each entry

Field	Type	Description
`rank`	`number`	Position among the companies this answer named — 1 means named first. The ORDER OF FIRST MENTION in `answerText`, computed by CompetLab from the text: the engine assigned no position, so never report it as a rank the engine gave. Stable under filtering.
`name`	`string`	The company’s name as the answer wrote it.
`domain`	`string`	The company’s primary domain, normalised — the identity.

item.answers[].sources[] — each entry

Field	Type	Description
`url`	`string`	Address of the retrieved page, after any redirect the engine returned was resolved.
`domain`	`string`	Domain of the retrieved page.
`title`	`string`	Title of the page as the engine reported it. Empty when the engine reported none.
`snippet`	`string`	The passage the engine handed back for this page beside its answer, verbatim. Present only where the engine reports one; absent means the engine reports none, never that the page has no text.
`publishedAt`	`string`	The page’s publication date as the engine reports it, verbatim — the engine’s string, not a measurement of ours. Present only where the engine reports one.
`resolvedFrom`	`string`	The address the engine actually returned, when `url` is the page it resolved to. Absent on a direct URL.

item.answers[].webSearch — object

Field	Type	Description
`performed`	`boolean`	Whether a live search RAN while the engine wrote this answer. `true` records that a search ran, not that the answer rests on what it found. For an engine read off a results page the search is the request, so it is true whenever the page came back.
`completedSearchCount`	`number`	How many searches the engine reports — each engine’s own quantity, and NOT comparable across engines: one reports billable requests, another only the query strings it lists. Compare an engine against itself over time; never compare engines with each other and never sum them. `0` beside `performed: false`.
`reason`	`string`	Present iff `performed` is false. `NO_SEARCH_ATTEMPTED` — the search tool was offered and went unused. `SEARCH_FAILED` — retrieval was attempted and did not succeed. `NO_SEARCH_REPORTED` — the answer carries no evidence of a search and the engine discloses no cause. None of these means the engine chose to answer from memory; they record that no retrieval is evidenced, and why is not something the data answers. Values: `NO_SEARCH_ATTEMPTED`, `SEARCH_FAILED`, `NO_SEARCH_REPORTED`.

item.engineStatus.perplexity — object

Field	Type	Description
`reported`	`boolean`	Whether this engine stored a result for this check. `false` means nothing came back from it at all. `true` does NOT mean it answered usefully — read the three counts beside it.
`completedAt`	`string \| null`	When this engine finished (ISO-8601). Null when it never reported. Nullable.
`questionsAsked`	`number`	Questions this engine was asked — the check’s own recorded question count.
`answersReceived`	`number`	Of `questionsAsked`: questions it answered — the entries for this engine in `answers`, before any filter. The three counts here are separate facts; state the shortfall against `questionsAsked` as a count, never as a ratio.
`answersAbsent`	`number`	Of `questionsAsked`: questions the engine was read for and showed no answer to — the entries for this engine in `noAnswerShown`. Not failures.
`answersUnmeasured`	`number`	Of `questionsAsked`: questions we could not read the answer to — the entries for this engine in `unansweredQueries`. Our problem, and in no count.

item.engineStatus.google_ai_overviews — object

Field	Type	Description
`reported`	`boolean`	Whether this engine stored a result for this check. `false` means nothing came back from it at all. `true` does NOT mean it answered usefully — read the three counts beside it.
`completedAt`	`string \| null`	When this engine finished (ISO-8601). Null when it never reported. Nullable.
`questionsAsked`	`number`	Questions this engine was asked — the check’s own recorded question count.
`answersReceived`	`number`	Of `questionsAsked`: questions it answered — the entries for this engine in `answers`, before any filter. The three counts here are separate facts; state the shortfall against `questionsAsked` as a count, never as a ratio.
`answersAbsent`	`number`	Of `questionsAsked`: questions the engine was read for and showed no answer to — the entries for this engine in `noAnswerShown`. Not failures.
`answersUnmeasured`	`number`	Of `questionsAsked`: questions we could not read the answer to — the entries for this engine in `unansweredQueries`. Our problem, and in no count.

item.summary.window.perEngine — object

Field	Type	Description
`perplexity`	`object`	Perplexity. Nested — see the Example tab.
`google_ai_overviews`	`object`	Google AI Overviews. Nested — see the Example tab.

item.summary.perEngine.perplexity — object

Field	Type	Description
`engineDataAvailable`	`object`	Present ONLY when this engine produced nothing usable this check. When present, no other field on this object is, and the card for this engine reads as unmeasured with this reason — never as zeros. Nested — see the Example tab.
`answersAsked`	`number`	Questions this engine was asked — the check’s own recorded question count, read off the check.
`answersReceived`	`number`	Questions this engine answered. THE denominator for every ‘answers’ figure on this engine. State the shortfall against `answersAsked` as two facts, never as a ratio.
`answersAbsent`	`number`	Questions the engine was read for and showed no answer to — today, questions Google’s results page carried no AI Overview for. Measured absences: not answers, not failures, and in no denominator.
`answersUnmeasured`	`number`	Questions we could not read this engine’s answer to. In no denominator. Our problem, never a fact about the question or the engine’s view of anyone.
`answersWithoutRetrieval`	`number`	Of `answersReceived`: answers the engine wrote without retrieving any page — it reported no page, and has not said how it answered; say `no pages reported`, never `from memory`. Still answers; they contribute no page.
`answersNamingCustomer`	`number`	Of `answersReceived`: answers that named the customer.
`answersRetrievingOwnPage`	`number`	Of `answersReceived`: answers in which the engine retrieved at least one page on the customer’s own site. Being read is not being recommended — compare with `answersNamingCustomer`.
`pagesRetrieved`	`number`	Distinct pages this engine RETRIEVED this check across its answers. Retrieved, never cited: the engine does not say which it leaned on. This engine’s number alone — never add it to another engine’s.
`hostsRetrieved`	`number`	Distinct hosts among `pagesRetrieved`.
`hostsOnlyThisEngine`	`number`	Of `hostsRetrieved`: hosts no other engine retrieved this check.
`engineCoreSize`	`number`	Hosts this engine retrieved for at least 2 of the questions — this engine’s own core, this check. The universe of that engine’s “seen in N of M” on the core hosts. Small on some engines by nature: quote it beside any “seen in” figure.
`pagesRead`	`number`	Of `pagesRetrieved`: pages whose body we read. The universe of `independentPagesNamingCustomer`.
`pagesUnreadable`	`number`	Of `pagesRetrieved`: pages we could not read. Listed, never counted — an unreadable page is never a page the customer is absent from.
`pagesNotFetched`	`number`	Of `pagesRetrieved`: pages whose fetch never reported. Kept apart from `pagesUnreadable` because ‘we stopped’ and ‘we could not read it’ are different facts.
`ownPagesRetrieved`	`number`	Of `pagesRetrieved`: pages on the customer’s own site, subdomains included. Never counted as independent.
`independentPagesNamingCustomer`	`object`	Independent pages this engine retrieved whose text names the customer — pages not on the customer’s own site — over `pagesRead`. A range: `floor` is the corroborated count, `ceiling` the naive one; equal for an unambiguous name. Named on a page the engine retrieved, not necessarily in the passage it read. Nested — see the Example tab.
`topHosts`	`array<object>`	The hosts this engine leaned on most this check, at most 10, by pages retrieved then host name. Each says which other engines also retrieved it. Nested — see the Example tab.

item.summary.perEngine.google_ai_overviews — object

Field	Type	Description
`engineDataAvailable`	`object`	Present ONLY when this engine produced nothing usable this check. When present, no other field on this object is, and the card for this engine reads as unmeasured with this reason — never as zeros. Nested — see the Example tab.
`answersAsked`	`number`	Questions this engine was asked — the check’s own recorded question count, read off the check.
`answersReceived`	`number`	Questions this engine answered. THE denominator for every ‘answers’ figure on this engine. State the shortfall against `answersAsked` as two facts, never as a ratio.
`answersAbsent`	`number`	Questions the engine was read for and showed no answer to — today, questions Google’s results page carried no AI Overview for. Measured absences: not answers, not failures, and in no denominator.
`answersUnmeasured`	`number`	Questions we could not read this engine’s answer to. In no denominator. Our problem, never a fact about the question or the engine’s view of anyone.
`answersWithoutRetrieval`	`number`	Of `answersReceived`: answers the engine wrote without retrieving any page — it reported no page, and has not said how it answered; say `no pages reported`, never `from memory`. Still answers; they contribute no page.
`answersNamingCustomer`	`number`	Of `answersReceived`: answers that named the customer.
`answersRetrievingOwnPage`	`number`	Of `answersReceived`: answers in which the engine retrieved at least one page on the customer’s own site. Being read is not being recommended — compare with `answersNamingCustomer`.
`pagesRetrieved`	`number`	Distinct pages this engine RETRIEVED this check across its answers. Retrieved, never cited: the engine does not say which it leaned on. This engine’s number alone — never add it to another engine’s.
`hostsRetrieved`	`number`	Distinct hosts among `pagesRetrieved`.
`hostsOnlyThisEngine`	`number`	Of `hostsRetrieved`: hosts no other engine retrieved this check.
`engineCoreSize`	`number`	Hosts this engine retrieved for at least 2 of the questions — this engine’s own core, this check. The universe of that engine’s “seen in N of M” on the core hosts. Small on some engines by nature: quote it beside any “seen in” figure.
`pagesRead`	`number`	Of `pagesRetrieved`: pages whose body we read. The universe of `independentPagesNamingCustomer`.
`pagesUnreadable`	`number`	Of `pagesRetrieved`: pages we could not read. Listed, never counted — an unreadable page is never a page the customer is absent from.
`pagesNotFetched`	`number`	Of `pagesRetrieved`: pages whose fetch never reported. Kept apart from `pagesUnreadable` because ‘we stopped’ and ‘we could not read it’ are different facts.
`ownPagesRetrieved`	`number`	Of `pagesRetrieved`: pages on the customer’s own site, subdomains included. Never counted as independent.
`independentPagesNamingCustomer`	`object`	Independent pages this engine retrieved whose text names the customer — pages not on the customer’s own site — over `pagesRead`. A range: `floor` is the corroborated count, `ceiling` the naive one; equal for an unambiguous name. Named on a page the engine retrieved, not necessarily in the passage it read. Nested — see the Example tab.
`topHosts`	`array<object>`	The hosts this engine leaned on most this check, at most 10, by pages retrieved then host name. Each says which other engines also retrieved it. Nested — see the Example tab.

item.summary.matrix[].cells — object

Field	Type	Description
`perplexity`	`object`	Perplexity’s cell for this question. Nested — see the Example tab.
`google_ai_overviews`	`object`	Google AI Overviews’ cell for this question. Nested — see the Example tab.

item.summary.brands[].answers — object

Field	Type	Description
`answersNaming`	`number`	Answers in this scope that named the brand — counted once per answer, matched on the brand’s domain, never on its wording.
`answersReceived`	`number`	The universe `answersNaming` is drawn from: answers received in this scope. Questions that produced no answer, or that the engine was read for and had nothing to show, are not in it. Quote it whenever you quote the share.
`presence`	`number`	`answersNaming / answersReceived × 100`, whole percent. A share of answers received — never of questions asked, and never a probability. Always read with `presenceLow` / `presenceHigh` beside it.
`presenceLow`	`number`	Lower bound of a 95% interval on `presence`, same percent scale. Two brands whose ranges overlap are NOT ordered, whatever their rank says.
`presenceHigh`	`number`	Upper bound of the same interval. See `presenceLow`.

item.summary.brands[].answersRetrievingOwnSite — object

Field	Type	Description
`answersNaming`	`number`	Answers in this scope that named the brand — counted once per answer, matched on the brand’s domain, never on its wording.
`answersReceived`	`number`	The universe `answersNaming` is drawn from: answers received in this scope. Questions that produced no answer, or that the engine was read for and had nothing to show, are not in it. Quote it whenever you quote the share.
`presence`	`number`	`answersNaming / answersReceived × 100`, whole percent. A share of answers received — never of questions asked, and never a probability. Always read with `presenceLow` / `presenceHigh` beside it.
`presenceLow`	`number`	Lower bound of a 95% interval on `presence`, same percent scale. Two brands whose ranges overlap are NOT ordered, whatever their rank says.
`presenceHigh`	`number`	Upper bound of the same interval. See `presenceLow`.

item.summary.brands[].perEngine — object

Field	Type	Description
`perplexity`	`object`	Perplexity’s reading of this brand. Nested — see the Example tab.
`google_ai_overviews`	`object`	Google AI Overviews’ reading of this brand. Nested — see the Example tab.

item.summary.overlap.perEngine — object

Field	Type	Description
`perplexity`	`object`	Perplexity. Nested — see the Example tab.
`google_ai_overviews`	`object`	Google AI Overviews. Nested — see the Example tab.

item.summary.coreHosts[].actionHint — object

Field	Type	Description
`code`	`string`	Stable identifier for this sentence. Branch on it if you need to; never substitute your own wording for it.
`text`	`string`	The sentence to report, in the words every CompetLab surface uses. Render it VERBATIM — do not paraphrase, shorten, or compose your own from the code. It names no row’s numbers on purpose; the fields beside it carry those.

item.summary.coreHosts[].seenInChecks — object

Field	Type	Description
`perplexity`	`object`	Perplexity. Nested — see the Example tab.
`google_ai_overviews`	`object`	Google AI Overviews. Nested — see the Example tab.

item.summary.coreHosts[].pages[] — each entry

Field	Type	Description
`key`	`string`	The page’s identity: the canonical form of its address — lower-case host without `www.`, no scheme, tracking parameters removed, no fragment. One page listed under two addresses is read once and counted once.
`url`	`string`	The address as an engine returned it.
`host`	`string`	The host, lower-case, without `www.`.
`title`	`string`	The page’s title — its own when we read it, else the title an engine reported. Absent when neither exists.
`engines`	`array<string>`	The engines that retrieved this page this check. Membership, never a count; its length is how many engines read it.
`promptIds`	`array<string>`	The questions whose answers retrieved this page, on any engine. Join on `matrix[].promptId`.
`ownedBy`	`string`	The domain of the brand whose own site this page is on, when it is on one — the customer’s or a competitor’s. Absent for a third-party page. A page on a brand’s own site is never an independent source naming that brand.
`readState`	`string`	`read` — the body was fetched and read. `unreadable` — we could not read it; `unreadableReason` says what was observed. `not_fetched` — no fetch ever reported for it. Kept apart from `unreadable` because ‘we stopped’ and ‘we could not read it’ are different facts. Values: `read`, `unreadable`, `not_fetched`.
`unreadableReason`	`string`	Present iff `readState` is `unreadable`. Each names what was OBSERVED about the fetch, never a fact about the page’s content: `bot_protection` — the site refused an automated reader; `host_blocked_recently` — not fetched, because the host refused us within the last month; `text_too_short` — a body arrived with too few readable characters; `not_found_page` — the server answered with a page that says the resource does not exist; `consent_wall` — the server answered with a consent interstitial in place of the page; `unreachable` — the host could not be reached; `timeout` — the page did not finish loading in time; `fetch_failed` — any other failure to obtain a body; `unresolved` — the engine returned a redirect we could not resolve to a page. Values: `bot_protection`, `host_blocked_recently`, `text_too_short`, `not_found_page`, `consent_wall`, `unreachable`, `timeout`, `fetch_failed`, `unresolved`.
`fetchedAt`	`string`	When this page’s verdict was taken (ISO-8601) — the age of the evidence. A page is re-read after a month. Absent when not fetched.
`textSource`	`string`	What text this page was judged on. `body` — the fetched page plus the title and passage the engine handed back. `snippet_only` — the title and passage alone, for a page we could not read. `none` — a page with neither. Values: `body`, `snippet_only`, `none`.
`namesCustomer`	`string \| null`	Whether this page’s text names the customer. `named` — it does, and where the name is ordinary language the page corroborates it. `named_uncorroborated` — the name matched but nothing corroborates it; reachable only for an ambiguous brand, and in the ceiling of every range rather than the floor. `not_named` — the page was READ and does not name the customer: the ONLY value that supports ‘get onto this page’. `null` — the page was not read (unreadable or not fetched), so nothing can be said. An unread page can still be `named` from the title and passage the engine handed back: an unread page can prove presence, never absence. Values: `named`, `named_uncorroborated`, `not_named`. Nullable.
`brandsNamed`	`array<string>`	Domains of the OTHER brands on the leaderboard that this page’s text names. Membership; the count is the array’s length. Empty is measured: we looked and found none.
`excerpt`	`string`	A short passage around the first mention of the customer where named, else of the first other brand named — at most 240 characters. Third-party text, quoted as evidence. Absent where no brand was found in the text.

item.summary.ownPageRetrievals[].pages[] — each entry

Field	Type	Description
`key`	`string`	The page’s identity — see `pages[].key`.
`url`	`string`	The address as the engine returned it.
`title`	`string`	The page’s title, when known.

item.summary.limits.unreadableByReason — object

Field	Type	Description
`bot_protection`	`number`	Pages a site refused to an automated reader.
`host_blocked_recently`	`number`	Pages recorded without a request, because their host refused us within the last month.
`text_too_short`	`number`	Pages whose body carried too few readable characters.
`not_found_page`	`number`	Pages the server answered with a page saying the resource does not exist. Listed, never counted.
`consent_wall`	`number`	Pages the server answered with a consent interstitial in place of the page. Listed, never counted.
`unreachable`	`number`	Pages whose host could not be reached.
`timeout`	`number`	Pages that did not finish loading in time.
`fetch_failed`	`number`	Pages that failed to fetch for any other reason.
`unresolved`	`number`	Redirect wrappers an engine returned that we could not resolve to a page.

item.summary.limits.textSourceMix — object

Field	Type	Description
`body`	`number`	Pages judged on the fetched body plus the engine’s title and passage.
`snippet_only`	`number`	Pages judged on the engine’s title and passage alone, because the body could not be read.
`none`	`number`	Pages with neither.

item.summary.limits.sentences[] — each entry

Field	Type	Description
`code`	`string`	Stable identifier for this sentence. Branch on it if you need to; never substitute your own wording for it.
`text`	`string`	The sentence to report, in the words every CompetLab surface uses. Render it VERBATIM — do not paraphrase, shorten, or compose your own from the code. It names no row’s numbers on purpose; the fields beside it carry those.

Errors

Status	Code	Meaning
400	`invalid_parameters` · `invalid_run_id` · `invalid_check_id`	Bad request — the payload failed validation.
401	`api_key_missing` · `api_key_invalid` · `api_key_revoked` · `api_key_expired` · `insufficient_scope`	The `CL-API-Key` is missing, malformed, revoked, expired, or lacks the required scope.
404	`check_not_found` · `run_not_summarized`	No `check` matches the id in the path.

Every error uses the shared { error: { code, message, status } } envelope; code is one of the values listed above.


{
  "error": {
    "code": "invalid_parameters",
    "message": "Domain is required",
    "status": 400
  }
}