> ## Documentation Index > Fetch the complete documentation index at: https://docs.peec.ai/llms.txt > Use this file to discover all available pages before exploring further. # Get AI Referrals > AI-referral traffic from the project's connected Google Analytics property, scoped to the hosts the project owns and to referrals from AI assistants. There is no way to ask this endpoint for non-AI traffic. To list the values a dimension takes, pass it as the only `group_by` and read the rows — that is how you discover event names, countries, pages and the assistants that actually sent traffic. When populating a filter, do not filter by the dimension you are listing, or the options the user has not already picked disappear. Metric definitions, because the names are shorter than the meanings: - `events`: GA4 event count. - `session_starts`: GA4 `session_start` events, which is not GA4's `sessions` metric. Most are visits that entered on the site; the rest resumed there after idling past GA4's session timeout. It cannot be combined with `group_by=event_name`. - `conversions`: GA4 key events, so one visit can contribute several, and which events count follows whatever the property marked as a key event. - `revenue`: GA4 purchase revenue in the property's reporting currency, returned as `currency`. Unknown dimension values come back as `(not set)`. It is a real row, and the free-string filters (`sources`, `devices`, `event_names`, `mediums`) accept it back; `countries` does not, because it takes ISO codes only. Figures are a lower bound on AI referrals: only referrer-visible traffic reaches GA4, so native apps and no-referrer policies land as Direct, and Gemini is often folded into Organic Search. ## OpenAPI ````yaml https://api.peec.ai/customer/v1/openapi/json get /agent-analytics/google-analytics/referrals openapi: 3.0.3 info: title: Peec AI Customer API description: Development documentation version: 1.0.0 contact: name: Peec AI Team email: support@peec.ai servers: - url: https://api.peec.ai/customer/v1 security: [] paths: /agent-analytics/google-analytics/referrals: get: tags: - Agent Analytics summary: Get AI Referrals description: >- AI-referral traffic from the project's connected Google Analytics property, scoped to the hosts the project owns and to referrals from AI assistants. There is no way to ask this endpoint for non-AI traffic. To list the values a dimension takes, pass it as the only `group_by` and read the rows — that is how you discover event names, countries, pages and the assistants that actually sent traffic. When populating a filter, do not filter by the dimension you are listing, or the options the user has not already picked disappear. Metric definitions, because the names are shorter than the meanings: - `events`: GA4 event count. - `session_starts`: GA4 `session_start` events, which is not GA4's `sessions` metric. Most are visits that entered on the site; the rest resumed there after idling past GA4's session timeout. It cannot be combined with `group_by=event_name`. - `conversions`: GA4 key events, so one visit can contribute several, and which events count follows whatever the property marked as a key event. - `revenue`: GA4 purchase revenue in the property's reporting currency, returned as `currency`. Unknown dimension values come back as `(not set)`. It is a real row, and the free-string filters (`sources`, `devices`, `event_names`, `mediums`) accept it back; `countries` does not, because it takes ISO codes only. Figures are a lower bound on AI referrals: only referrer-visible traffic reaches GA4, so native apps and no-referrer policies land as Direct, and Gemini is often folded into Organic Search. operationId: getAgent-analyticsGoogle-analyticsReferrals parameters: - name: project_id in: query required: false schema: description: Required if using a company api key example: or_f45b94ba-5e35-4982-93ed-285e72ee14eb type: string - name: start_date in: query required: true schema: type: string format: date pattern: >- ^(?:(?:\d\d[2468][048]|\d\d[13579][26]|\d\d0[48]|[02468][048]00|[13579][26]00)-02-29|\d{4}-(?:(?:0[13578]|1[02])-(?:0[1-9]|[12]\d|3[01])|(?:0[469]|11)-(?:0[1-9]|[12]\d|30)|(?:02)-(?:0[1-9]|1\d|2[0-8])))$ description: >- Inclusive bound, as full-date notation per RFC 3339 section 5.6. Required. A range spans at most 366 days. example: '2026-09-22' - name: end_date in: query required: true schema: type: string format: date pattern: >- ^(?:(?:\d\d[2468][048]|\d\d[13579][26]|\d\d0[48]|[02468][048]00|[13579][26]00)-02-29|\d{4}-(?:(?:0[13578]|1[02])-(?:0[1-9]|[12]\d|3[01])|(?:0[469]|11)-(?:0[1-9]|[12]\d|30)|(?:02)-(?:0[1-9]|1\d|2[0-8])))$ description: >- Inclusive bound, as full-date notation per RFC 3339 section 5.6. Required. A range spans at most 366 days. example: '2026-09-22' - name: group_by in: query required: false schema: description: >- Dimension(s) to break the numbers down by, at most 3. Omit for period totals only. Pass a single dimension to list the values it takes. example: - assistant type: array items: type: string enum: - assistant - platform - source - medium - country - device - event_name - landing_page - page_path - host - name: metrics in: query required: false schema: description: >- Metrics to return. Defaults to all of them, minus session_starts when grouping by event_name. Unrequested metrics are left out of the query. example: - session_starts - conversions type: array items: type: string enum: - events - session_starts - conversions - revenue - name: time_bucket in: query required: false schema: description: >- Adds a time_bucket column holding the bucket's start date. There is no hourly bucket: the source data is daily. Buckets at either end of the range cover only the days inside it. example: week type: string enum: - day - week - month - name: assistants in: query required: false schema: description: >- Filter by assistant display name. An unrecognised name is rejected rather than ignored. example: - ChatGPT type: array items: type: string - name: sources in: query required: false schema: description: Filter by raw GA4 sessionSource host. example: - chatgpt.com type: array items: type: string - name: countries in: query required: false schema: description: Filter by ISO 3166-1 alpha-2 country code. example: - DE type: array items: type: string enum: - AE - AL - AM - AR - AT - AU - BA - BE - BG - BH - BO - BR - BS - BY - CA - CH - CL - CN - CO - CR - CY - CZ - DE - DK - DO - EC - EE - EG - ES - FI - FR - GB - GE - GH - GR - GT - HK - HN - HR - HU - ID - IE - IL - IN - IQ - IS - IT - JO - JP - KR - KW - LB - LI - LT - LU - LV - MA - MD - ME - MK - MN - MT - MX - MY - NG - NI - NL - 'NO' - NZ - OM - PK - PA - PE - PH - PL - PT - PY - PS - QA - RO - RS - SA - SE - SG - SI - SK - SV - TH - TN - TR - TW - UA - US - UY - VE - VN - ZA - AD - AF - AS - AZ - BB - BQ - CG - CI - CM - CW - DM - DZ - FO - GF - GP - JM - KG - KH - KI - KZ - LK - LR - LS - MW - NC - PG - TD - TF - UG - VU - ZW - name: mediums in: query required: false schema: description: >- Filter by GA4 session medium. Use `ai-assistant` to isolate traffic identified by medium rather than by a known referrer host. example: - ai-assistant type: array items: type: string - name: devices in: query required: false schema: description: >- Filter by device category. Property-specific and case-sensitive: an unmatched value yields zeros rather than an error. example: - mobile type: array items: type: string - name: event_names in: query required: false schema: description: >- Filter by GA4 event name. Property-specific: a name this property never sends yields zeros rather than an error, so discover them with `group_by=event_name` first. example: - purchase type: array items: type: string - name: order_by in: query required: false schema: description: >- Field to sort by, which must be one of the requested metrics (or time_bucket when it is set). Defaults to the first requested metric. example: session_starts type: string enum: - events - session_starts - conversions - revenue - time_bucket - name: order_direction in: query required: false schema: default: desc type: string enum: - asc - desc - name: include_total_count in: query required: false schema: description: >- Set false to skip total_count. Counting is a second pass over the same rows, so it doubles the cost of the call. example: 'false' default: 'true' type: string enum: - 'true' - 'false' - name: limit in: query required: false schema: default: 100 type: integer minimum: 1 maximum: 1000 - name: offset in: query required: false schema: default: 0 type: integer minimum: 0 maximum: 9007199254740991 responses: '200': description: Success content: application/json: schema: type: object properties: data: type: array items: type: object properties: assistant: example: ChatGPT type: string platform: type: string example: OpenAI source: example: chatgpt.com type: string medium: example: referral type: string country: example: DE type: string device: example: mobile type: string event_name: example: page_view type: string landing_page: description: >- The session's entry page, repeated on every event in that session. Query string stripped. example: /pricing type: string page_path: description: >- The page the event fired on. Query string stripped and UUIDs collapsed to `*`, so it does not join to landing_page. example: /blog/*/overview type: string host: example: www.example.com type: string time_bucket: description: >- First day of the bucket this row covers, always a plain date. Present only when the `time_bucket` request parameter was set, which names the width rather than a date. example: '2026-03-01' type: string events: example: 4820 type: number session_starts: type: number example: 1402 conversions: example: 37 type: number revenue: example: 4210.5 type: number columns: type: array items: type: string description: >- Column order for the rows, derived from the request so a zero-row answer still carries its shape. example: - assistant - session_starts total_count: nullable: true description: >- Number of grouped rows the filters match — not a sum of any metric. Null when include_total_count was false. example: 12 type: number currency: nullable: true description: >- Reporting currency of `revenue`, null when revenue was not requested. Values are not converted, so revenue is not comparable across projects. example: EUR type: string required: - data - columns - total_count - currency description: Success '409': description: Response for status 409 content: application/json: schema: type: object properties: message: type: string required: - message security: - APIKeyHeader: [] - APIKeyQuery: [] - BearerAuth: [] components: securitySchemes: APIKeyHeader: type: apiKey in: header name: X-API-Key APIKeyQuery: type: apiKey in: query name: api_key BearerAuth: type: http scheme: bearer bearerFormat: JWT ````