GEO Monitoring API: Track Brand Mentions in AI Answers

You create a monitor once. Each run is a set of ordinary async tasks, billed at each engine’s price. This page covers creating, running and managing monitors. Reading the numbers is in Reports and metrics, the answers behind them in Sources and answers, and webhooks and email alerts in Events and alerts. For the dashboard view of the same data, see the product page.

Brand and category monitors

A monitor has a mode, set when you create it and fixed for its life, because it decides what every stored answer means.

Mode	Follows	You provide	The report shows
`BRAND` (default)	One brand and its competitors	`aliases` (your brand’s names), optionally `domains` and `competitors`	Your mention rate, citation rate, share of voice and the prompts where competitors appear without you
`CATEGORY`	A market	`category`, optionally `promptTopics` (sub-fields) and `competitors`	A leaderboard of the brands the answers name, per sub-field and engine

Each answer is scored once, when it arrives:

Mentioned: one of your aliases appears in the answer text. position is the character offset of the first match, or null when the brand is absent.
Cited: the answer cites a page on one of your domains or a subdomain of it.
Competitor hits: which of the monitor’s competitors appear in the same answer.

These scores carry no sentiment and no estimate of market share. A category monitor’s rank is a place among the brands the answers name, nothing more. For sentiment and recommendations per answer, turn on analysis.

Quick start

Create the monitor

The first run starts as soon as the monitor is created.

curl -X POST https://api.querying.ai/v1/monitors \
  -H "Authorization: Bearer $QUERYING_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "name": "Earbud brand tracking",
    "engines": ["CHATGPT", "GEMINI"],
    "prompts": ["best wireless earbuds for commuting", "are cheap earbuds worth it"],
    "aliases": ["Acme Audio", "Acme"],
    "domains": ["acme-audio.example"],
    "competitors": [{ "name": "Northwind Audio" }, { "name": "Contoso Sound", "aliases": ["Contoso"] }],
    "country": "US",
    "intervalHours": 24
  }'

The response is 201 with the stored monitor. Keep data.monitor.id.

Wait for the answers

Tasks are admitted as your plan’s concurrency allows, so a run arrives over seconds to minutes. The report’s health.pending counts the tasks still in flight. Read the report once it reaches 0.

Read the report

curl "https://api.querying.ai/v1/monitors/$MONITOR_ID/analytics?days=7" \
  -H "Authorization: Bearer $QUERYING_API_KEY"

Read quality first, then stats, brands and opportunities. Every field is explained in Reports and metrics.

To get answers before the next scheduled run, call POST /v1/monitors/{id}/run.

Create a monitor

POST /v1/monitors (reference) validates the definition, checks it against your account’s limits, stores it and starts the first run. The call returns 201 before that run finishes. If the first run cannot start (for example, your balance is too low), the monitor is still created and the scheduler retries the run every minute. For example, the quick start request returns:

{
  "success": true,
  "data": {
    "monitor": {
      "id": "d39b8de0-b159-4229-bf58-7064e5369db1",
      "name": "Earbud brand tracking",
      "mode": "BRAND",
      "category": null,
      "promptTopics": {},
      "engines": ["CHATGPT", "GEMINI"],
      "prompts": ["best wireless earbuds for commuting", "are cheap earbuds worth it"],
      "aliases": ["Acme Audio", "Acme"],
      "domains": ["acme-audio.example"],
      "country": "US",
      "intervalHours": 24,
      "enabled": true,
      "alertBelowPct": null,
      "competitors": [
        { "name": "Northwind Audio", "aliases": ["Northwind Audio"], "source": "user" },
        { "name": "Contoso Sound", "aliases": ["Contoso Sound", "Contoso"], "source": "user" }
      ],
      "competitorsReadAt": null,
      "location": null,
      "grid": null,
      "engineSettings": {},
      "webhookUrl": null,
      "analysis": false,
      "nextRunAt": "2026-10-20T09:00:04.512Z",
      "lastRunAt": "2026-10-19T09:00:04.512Z",
      "shareAccess": "PRIVATE",
      "shareSlug": null,
      "createdAt": "2026-10-19T09:00:03.977Z",
      "updatedAt": "2026-10-19T09:00:03.977Z"
    }
  }
}

Request fields

Every list field accepts a JSON array or one newline-separated string. Entries are trimmed, blank entries are dropped and duplicates are removed.

Field	Required	Description
`name`	yes	1–80 characters.
`mode`	no	`BRAND` (default) or `CATEGORY`. Cannot change after creation.
`engines`	yes	The engines to ask. See Engines.
`prompts`	yes	The questions to ask, each up to 2,000 characters. Up to 100 per monitor, or 1,000 on a contract.
`aliases`	brand: yes	Up to 20 names that mean your brand. Each needs at least 2 characters and cannot be a URL. A category monitor rejects a non-empty value.
`domains`	no	Up to 20 sites you own, such as `acme-audio.example`. A URL is stored as its host without `www.`. Without domains, citation tracking is off and `cited` is always `false`. A category monitor rejects a non-empty value.
`competitors`	no	Brands to compare against: `{ "name": "...", "aliases": ["..."] }` objects or `"Name: alias, alias"` strings. Up to 10 on a brand monitor, 25 on a category monitor. A name is at most 80 characters and has up to 20 aliases; `__you__` is reserved. See Competitors.
`category`	category: yes	The market, as a buyer names it: 1–80 characters. A brand monitor rejects it.
`promptTopics`	no	Category only. Maps a prompt’s exact text to its sub-field, for example `{ "best sunscreen for kids": "Kids" }`. Up to 30 sub-fields of at most 60 characters. Every key must be one of `prompts`. A prompt without an entry covers the whole category.
`country`	no	ISO 3166-1 alpha-2 code of the market the engines answer for. Default `US`. It does not translate your prompts.
`intervalHours`	yes	Hours between runs: a whole number from 1 to 168.
`engineSettings`	no	Per-engine answers per prompt and cadence. See Samples and per-engine cadence.
`location`	no	`{ "latitude": 40.7128, "longitude": -74.006 }`: every task asks from this point. Cannot be combined with `grid`. See Location and grids.
`grid`	no	`{ "center": { "latitude": ..., "longitude": ... }, "size": 5, "radiusMeters": 2000 }`: every prompt is asked from each point of a grid. Cannot be combined with `location`.
`alertBelowPct`	no	Brand only. A whole percentage from 1 to 100: email the account when the 7-day mention rate falls below it. `null` turns it off. See Alerts.
`webhookUrl`	no	An HTTPS URL on a public host that receives change events, or `null`. Set it with an API key: events are signed with the key that sets it.
`analysis`	no	`true` asks every task for answer analysis with the monitor’s brand, domains and competitors. Default `false`.

Engines

A monitor can schedule the prompt-based answer engines: CHATGPT, GEMINI, PERPLEXITY, GOOGLE (Google AI Overviews), AIMODE (Google AI Mode), NAVER_AI_BRIEF, NAVER_AI_TAB, and NAVER, a deprecated name kept for existing monitors. The aliases GOOGLE_AIO, GOOGLE_AIMODE and NAVER_AI_MODE are accepted and stored as GOOGLE, AIMODE and NAVER_AI_TAB. Engine names are case-sensitive. Search, web data and analysis task types, such as GOOGLE_SERP or CITATION_ANALYSIS, cannot be scheduled. Prices and per-engine details are in the engine catalog. An API key restricted to some engines can only create, update or run monitors that use those engines; anything else returns 403 KEY_SCOPE_DENIED.

Competitors

You do not have to list competitors. After the first run, a weekly read of the monitor’s recent answers adds the brands that sell what you sell and appear in at least two answers: up to 15 on a brand monitor, 25 on a category monitor.

Found competitors carry source: "auto"; the ones you list carry source: "user". The read never changes or removes yours. A found brand that the next two reads both miss is removed.
When the found list changes, the monitor’s stored answers are scored again against it, so every brand in a report is measured over the same answers. competitorsReadAt says when the last read ran.
On update, competitors replaces your own list and keeps the found ones. Entries with source: "auto" in the request are ignored, so you can send back what GET returned.

How names are matched

Matching ignores case and character width (NFKC).
Names in Latin script match whole words only: “replicates” is not a mention of a brand called Replicate. A Korean, Japanese, Chinese or Thai letter next to the name counts as a word break, so “agoda를” and “ahamoが” mention agoda and ahamo. Korean, Japanese, Chinese and Thai names match anywhere in the text.
A one-character name never matches. Give such a brand a longer alias.
When two names overlap, the one that starts first, then the longer one, counts. “Acme” inside a competitor named “Acme Pro” is the competitor.
Link targets, bare URLs and citation chips are removed before matching; the visible text of a link is kept. A citation counts toward cited, not toward a mention.

Write prompts that measure something

One monitor per topic and market. “best CRM” and “how do I price a CRM” are different intentions; averaging them hides both.
Phrase prompts as a buyer would, and never include your own brand. A prompt that names the brand is always a mention, so it reports 100% and measures nothing. Competitor names in prompts are fine.

Samples and per-engine cadence

AI answers change from one session to the next, so a share of several answers is steadier than one. engineSettings sets, per engine, how many answers to collect per prompt and how often the engine runs:

{
  "intervalHours": 24,
  "engineSettings": {
    "CHATGPT": { "samples": 5 },
    "GOOGLE": { "intervalHours": 168 }
  }
}

samples: 1–10 answers per prompt and point in each run (default 1). Each is a separate task in its own session, scored and billed like any other.
intervalHours: a multiple of the monitor’s intervalHours, up to 168. The engine joins every k-th run, counting from the first run, which every engine joins.

An entry can only name one of the monitor’s engines and accepts only these two keys. On update, engineSettings replaces the stored value; an update that drops an engine drops its entry, and one that changes the monitor’s intervalHours checks the engine intervals against the new value.

Location and grids

By default every task asks for the country as a whole. To measure local answers:

location asks every prompt from one point, in decimal degrees.
grid asks every prompt from each point of a square grid around center. size is the points per side: 3, 5, 7 or 9 (9 to 81 points). radiusMeters, 100 to 50,000, is the distance from the center to the middle of an edge.

Only CHATGPT, GEMINI, GOOGLE and AIMODE answer for a location. A monitor with a location or grid and any other engine is rejected with 400 VALIDATION_ERROR. On update, setting one of location and grid clears the other, and null clears it. Read the answers per point with GET /v1/monitors/{id}/grid.

Cost and scheduling

A run creates, for each engine that joins it, one task per prompt, per point and per sample:

tasks per run    = sum over engines in the run of (prompts × points × samples)
credits per run  = sum over engines in the run of (prompts × points × samples × engine credits)
credits / month  ≈ credits per run × 720 / intervalHours

points is 1 without a grid. For example, 12 prompts on CHATGPT (1 credit) and GOOGLE (1.25 credits) is 24 tasks and 27 credits per run, about 810 credits a month at a daily interval. With "CHATGPT": { "samples": 5 } the same monitor is 72 tasks and 75 credits per run. Read current prices from GET /v1/monitors/capabilities rather than hardcoding them; how credits are held and charged is in Credits and limits. Two charges come on top of the tasks:

Competitor read: 80 credits per 30 days per monitor, split across its runs by cadence (about 2.67 credits per run for a daily monitor, 18.67 for a weekly one). It is charged when a run’s tasks are all submitted, so a run your balance stopped costs nothing for it.
Answer analysis: with analysis: true, 1 credit per task, charged only for answers that were analyzed. The scores (mentioned, cited, position) are unchanged; each answer’s result is in answer.analysis from GET /v1/monitors/{id}/answers/{taskId}.

Reading reports, sources, answers and results costs nothing.

The schedule repeats until you pause or delete the monitor. The monthly figure above is what you commit to when you create it.

How the schedule behaves:

A run is not a burst. Tasks are submitted as your plan’s concurrency and the engine queues allow. Every minute the scheduler continues where it stopped, one task per engine in turn, and monitors of different accounts take turns. A parent account and its sub-accounts share one set of turns.
A run must fit its interval. If one engine’s tasks cannot be admitted before that engine runs again, create and update return 400 VALIDATION_ERROR naming the shortest interval that fits. A run still submitting when a full interval has passed is closed, and its remaining tasks are not sent.
Missed runs are not replayed. nextRunAt is when the next run becomes due. Resuming a paused monitor, or changing intervalHours, schedules the next run one interval from that moment.

Capabilities

GET /v1/monitors/capabilities (reference) returns, for your account: each schedulable engine with creditsPerTask, analysisCreditsPerTask, your limits, the modes, the metric definitions (metrics), the schedule formulas and the change event types. It has no side effects and costs nothing; call it before creating monitors programmatically.

curl https://api.querying.ai/v1/monitors/capabilities \
  -H "Authorization: Bearer $QUERYING_API_KEY"

Choose prompts

Suggest prompts from brand facts

POST /v1/monitors/suggest (reference) returns up to 12 candidate prompts written as a buyer in country would ask. It saves nothing and spends no credits.

Field	Required	Description
`name`	one of `name`, `aliases`	Your brand. Defaults to the first alias.
`aliases`	one of `name`, `aliases`	Other names of the brand. Up to 20 are read.
`domains`	no	Your sites, as context.
`competitors`	no	Competitor names, as context. `"Name: alias"` strings are read as the name.
`country`	no	Two-letter market code. Default `US`.

curl -X POST https://api.querying.ai/v1/monitors/suggest \
  -H "Authorization: Bearer $QUERYING_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{ "name": "Acme Audio", "domains": ["acme-audio.example"], "competitors": ["Northwind Audio"], "country": "US" }'

For example:

{
  "success": true,
  "data": {
    "prompts": [
      "best wireless earbuds for commuting on a train",
      "noise cancelling earbuds under $100",
      "are budget earbuds worth it compared to premium ones",
      "how to choose earbuds for running"
    ]
  }
}

Suggestions that contain the brand name or an alias are removed. Review the list before you create the monitor. You can request one suggestion every 20 seconds and 30 per hour; more returns 429 RATE_LIMITED. When the model declines the brand the call returns 400 SUGGEST_FAILED, when the suggestion service fails 502 SUGGEST_FAILED, and when it is not configured 503 SUGGEST_FAILED.

Research prompts from search demand

POST /v1/monitors/research (reference) takes the request body of Prompt Research, costs the same 11 credits and returns the same result as a task owned by your account.

curl -X POST https://api.querying.ai/v1/monitors/research \
  -H "Authorization: Bearer $QUERYING_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{ "seed": "sunscreen", "country": "US", "monitorSize": 10, "monitorEngines": ["CHATGPT", "GEMINI"] }'

The response carries the queued task in data.task, with data.credits and data.replayed. Read the result with GET /v1/async/task/{id} using data.task.id (see Async tasks). Then create the monitor with monitorSet.prompts[].prompt as prompts; on a category monitor, each prompt’s topic as its entry in promptTopics and brands[].brand as competitors. Nothing is saved until you create the monitor. With monitorEngines, research checks monitorSize × engines against a fixed 200 tasks per run and assumes one point and one sample per engine, so a set it accepts can still be refused once you add samples or a grid. See Sizing the set for a monitor.

Category monitors

A category monitor ranks the brands the engines name for a market. Group its prompts by sub-field with promptTopics:

curl -X POST https://api.querying.ai/v1/monitors \
  -H "Authorization: Bearer $QUERYING_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "name": "Sunscreen market",
    "mode": "CATEGORY",
    "category": "Sunscreen",
    "engines": ["CHATGPT", "GEMINI"],
    "prompts": ["best sunscreen for sensitive skin", "best sunscreen for kids"],
    "promptTopics": {
      "best sunscreen for sensitive skin": "Sensitive skin",
      "best sunscreen for kids": "Kids"
    },
    "competitors": [{ "name": "Northwind Sun" }, { "name": "Contoso Skin", "aliases": ["Contoso"] }],
    "country": "US",
    "intervalHours": 24
  }'

A category monitor has no brand of its own: aliases, domains and alertBelowPct must be empty, or the request returns 400 VALIDATION_ERROR, and it has no alert. Its stats.mentioned and stats.cited in GET /v1/monitors are always 0; the list adds leader, the brand named most in the last 30 days. Its report is described in Category reports. Schedule, cost, sources, answers, results and change events work as on a brand monitor, except that a category monitor sends only monitor.source.first_cited events. On update, an edit to prompts drops the promptTopics entries of removed prompts. The answers those prompts collected stay in the leaderboard.

Manage monitors

Method and path	What it does
`GET /v1/monitors`	Your monitors, newest first, each with `stats` (`runs`, `mentioned`, `cited`) over the last 30 days. `data.windowDays` is `30`.
`GET /v1/monitors/{id}`	One monitor’s stored definition, with an older period summary.
`PATCH /v1/monitors/{id}`	Update any create field except `mode`, or send `enabled` to pause and resume.
`DELETE /v1/monitors/{id}`	Delete the monitor and every scored answer it holds. Returns `data.deleted: true`.
`POST /v1/monitors/{id}/run`	Make the next run due now.

The older period summary in GET /v1/monitors/{id} is kept for compatibility. For numbers, use /analytics: the older summary mixes an all-time matrix of the latest answer per prompt and engine into windowed numbers, cuts sources at 12 domains and 20 pages, and its shareOfVoice counts each brand against all answers, so the shares can add up to more than 100. A monitor that belongs to another account returns 404 NOT_FOUND, not 403.

Update a monitor

PATCH takes any subset of the create fields and validates the result as a whole, against the same limits. Fields you omit keep their values.

curl -X PATCH https://api.querying.ai/v1/monitors/$MONITOR_ID \
  -H "Authorization: Bearer $QUERYING_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{ "enabled": false }'

enabled: false pauses the monitor; enabled: true resumes it with the next run one interval from now. Pauses appear in reports as pauses.
Changing alertBelowPct re-arms the alert.
Changing aliases, domains or competitors affects answers scored from then on. Stored answers keep the definitions they were scored with, so reports withhold comparisons across the change (see Reports and metrics).

Run a monitor now

POST /v1/monitors/{id}/run makes the next run due immediately; the scheduler starts it within about a minute and returns data.queued: true. The run is charged like a scheduled one. Calling it again while that run is still due does not create a second run. Before queueing, the call checks that your balance covers the run, including the competitor read and analysis credits. If it does not, it returns 402 INSUFFICIENT_CREDITS with details.needed and details.available. A sub-account over its monthly credit limit gets 429 SUB_ACCOUNT_CREDIT_LIMIT. A paused monitor returns 404 NOT_FOUND: resume it first. You can publish one monitor’s report at https://querying.ai/<slug>. A shared report shows the analytics, sources, citations, results and answers; it never shows settings, alerts, the run controls or your account’s task health. GET /v1/monitors/{id}/share (reference) returns the current state, access, slug, locale and url, plus a suggestedSlug when none is set. PUT /v1/monitors/{id}/share (reference) changes it:

curl -X PUT https://api.querying.ai/v1/monitors/$MONITOR_ID/share \
  -H "Authorization: Bearer $QUERYING_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{ "access": "LINK", "slug": "acme-earbuds", "locale": "en" }'

Field	Required	Description
`access`	yes	`PRIVATE` (only you), `LINK` (anyone with the URL; not indexed) or `PUBLIC` (anyone; search engines may index it).
`slug`	the first time	3–40 lowercase letters, digits or single hyphens, starting and ending with a letter or digit. Some words are reserved for the site’s own pages. A slug in use returns `409 SLUG_TAKEN`.
`locale`	no	The language the report renders in for every reader: `en`, `ko`, `ja`, `zh`, `es`, `fr`, `de` or `pt`.

Setting access to PRIVATE takes the report offline and keeps the slug, so turning sharing on again reuses the same address.

Use monitors from an agent

The MCP server exposes monitors as tools: get_monitor_capabilities, list_monitors, get_monitor, create_monitor, update_monitor, delete_monitor, run_monitor, get_monitor_analytics, get_monitor_sources, get_monitor_results, get_monitor_answer, suggest_monitor_prompts, get_monitor_alerts and test_monitor_alert. Each makes one API call with your key. A practical sequence for an agent or a script:

GET /v1/monitors/capabilities: engines, prices and limits before spending anything.
GET /v1/monitors: reuse a monitor for the topic and market, or create one, optionally starting from POST /v1/monitors/suggest.
POST /v1/monitors/{id}/run only if you need answers before the next scheduled run.
GET /v1/monitors/{id}/analytics?days=30: read quality before the numbers, and wait until health.pending reaches 0.
GET /v1/monitors/{id}/sources?days=30&groupBy=page: follow nextCursor to the end.
GET /v1/monitors/{id}/answers/{taskId}: open the answer behind the top entry in opportunities before acting on it.
GET /v1/monitors/{id}/results?since=...&until=...&format=csv: export the same window, following x-next-cursor until it is empty.

Limits

Limit	Standard	Contract
Monitors per account (a parent and its sub-accounts together)	20	10,000
Prompts per monitor	100	1,000
Tasks per run (prompts × points × samples, summed over engines)	200	100,000

A contract account is one with a negotiated concurrency limit. These apply to every account:

Limit	Value
Prompt length	2,000 characters
Monitor name	80 characters
Brand aliases	20
Registered domains	20
Competitors you list	10 on a brand monitor, 25 on a category monitor
Competitors found by the weekly read	up to 15 on a brand monitor, 25 on a category monitor
Category name	80 characters
Sub-fields per category monitor	30, each up to 60 characters
Samples per prompt and engine	1–10
Interval	1–168 hours
Engine interval	a multiple of the monitor’s interval, up to 168 hours
Grid	3, 5, 7 or 9 points per side; radius 100–50,000 meters

Report, page-size and retention limits are listed on the pages that use them: Reports and metrics, Sources and answers and Events and alerts.

Errors

The error envelope is the same as everywhere else (see Errors). Codes specific to monitors:

Status	Code	When
400	`VALIDATION_ERROR`	An invalid definition (unknown engine, a list over its limit, `mode` changed, fields that do not belong to the mode, a location with an engine that cannot use one), a run over the tasks-per-run limit or too large for its interval, `webhookUrl` set without an API key, or invalid query parameters on a read. The message names the problem.
400	`SUGGEST_FAILED`	The suggestion model declined to write prompts for this brand.
401	`MISSING_API_KEY`, `UNAUTHORIZED`	No key, or an invalid or revoked key.
402	`INSUFFICIENT_CREDITS`	`POST /run` when your balance does not cover one run.
403	`KEY_SCOPE_DENIED`	The key’s allowed engines do not cover the monitor’s engines.
404	`NOT_FOUND`	No such monitor for this account, a paused monitor on `POST /run`, or no scored answer or event with that id.
409	`MONITOR_LIMIT`	The account (with its parent and sibling sub-accounts) already holds the maximum number of monitors.
409	`SLUG_TAKEN`	The share slug belongs to another report.
409	`WEBHOOK_NOT_SET`	Redelivering an event on a monitor without `webhookUrl`.
429	`RATE_LIMITED`	Prompt suggestions (one per 20 seconds, 30 per hour) or test alerts (three per monitor per five minutes).
429	`SUB_ACCOUNT_CREDIT_LIMIT`	`POST /run` from a sub-account over its monthly credit limit.
502, 503	`SUGGEST_FAILED`	The suggestion service failed, or is not configured.

Next steps

Reports and metrics

The /analytics contract and what each number means.

Sources and answers

Cited pages, the answers behind each score, and exports.

Events and alerts

Webhooks when answers change, and email alerts on your mention rate.

Prompt Research

Find the prompts worth tracking from search demand.

Get started

Core concepts

Answer engines

Search and web data

Request options

Research

GEO monitoring

Account and data

GEO Monitoring API: Track Brand Mentions in AI Answers

Brand and category monitors

Quick start

Create a monitor

Request fields

Engines

Competitors

How names are matched

Write prompts that measure something

Samples and per-engine cadence

Location and grids

Cost and scheduling

Capabilities

Choose prompts

Suggest prompts from brand facts

Research prompts from search demand

Category monitors

Manage monitors

Update a monitor

Run a monitor now

Use monitors from an agent

Limits

Errors

Next steps

Reports and metrics

Sources and answers

Events and alerts

Prompt Research

​Brand and category monitors

​Quick start

​Create a monitor

​Request fields

​Engines

​Competitors

​How names are matched

​Write prompts that measure something

​Samples and per-engine cadence

​Location and grids

​Cost and scheduling

​Capabilities

​Choose prompts

​Suggest prompts from brand facts

​Research prompts from search demand

​Category monitors

​Manage monitors

​Update a monitor

​Run a monitor now

​Share a report

​Use monitors from an agent

​Limits

​Errors

​Next steps

Reports and metrics

Sources and answers

Events and alerts

Prompt Research

Brand and category monitors

Quick start

Create a monitor

Request fields

Engines

Competitors

How names are matched

Write prompts that measure something

Samples and per-engine cadence

Location and grids

Cost and scheduling

Capabilities

Choose prompts

Suggest prompts from brand facts

Research prompts from search demand

Category monitors

Manage monitors

Update a monitor

Run a monitor now

Share a report

Use monitors from an agent

Limits

Errors

Next steps