> ## Documentation Index
> Fetch the complete documentation index at: https://api-docs.hrizn.io/llms.txt
> Use this file to discover all available pages before exploring further.

# Search and GA4 analytics

> Read warehoused Search Console and GA4 through the Public API and MCP

Use the analytics endpoints to **notice** search movers, **diagnose** a URL (which queries it ranks for, and whether it is indexed), then **create** content. Search Analytics numbers come from the warehouse. The only live Google read is URL Inspection, and it runs only on page performance.

## Who can call this

* **Public API:** `analytics:read` plus **API Access** (Teams V2 or higher). Existing keys do not pick up the new scope automatically — add `analytics:read` in the dashboard or create a key from a Read / MCP preset.
* **MCP:** granted rooftops need **Teams or higher** with API Access. Tools use the same warehouse as REST.

`GOOGLE_SERVICES` is required for Search Console and GA4 warehouse reads. A missing plan is `402` / upgrade-required. A rooftop that has never connected Search Console is `connected: false` — do not treat an API Access denial as disconnected.

## Notice → diagnose → create

1. `GET /analytics/search/snapshot` and `GET /analytics/search/alerts` (MCP: `get_search_snapshot`, `get_search_alerts`, or `get_dealer_briefing`).
2. `GET /analytics/search/opportunities` and `GET /analytics/search/gaps` for what to write. Keep `GET /content-intelligence` for nightly Content Intelligence recommendations — those are a different pipeline.
3. `GET /analytics/pages/performance?page_url=…` (MCP: `get_page_performance`) for query×page plus URL Inspection (24h cache).
4. `GET /analytics/search/create-brief`, then create with IdeaClouds or Content.

Pulse (`GET /analytics/search/pulse`) is REST-only. Group site lists can include pulse; there is no MCP pulse tool.

## Warehouse vs live Google

| Data                                                                                                  | Source                                            | Freshness                                                               |
| ----------------------------------------------------------------------------------------------------- | ------------------------------------------------- | ----------------------------------------------------------------------- |
| Search queries, pages, opportunities, gaps, changes, decay, questions, alerts, snapshot, create-brief | Search Console warehouse (web + image query×page) | Nightly GSC sync. Payload `as_of` is the last completed warehouse date. |
| Traffic snapshot and pages                                                                            | GA4 warehouse                                     | Nightly GA4 sync                                                        |
| URL Inspection (`index_status`, crawl, last crawl)                                                    | Live Google URL Inspection API                    | Only on `GET /analytics/pages/performance`. Results cached 24 hours.    |

Do not call Search Analytics API v2 from this product. Listing endpoints do not inspect URLs. `GET /content/{id}` may attach a **cached** inspection strip; it never triggers a live inspect.

## 50k query×page cap

Each GSC connection **syncs** up to **50,000** query×page rows per search type (`web` and `image`) per day. Snapshot, query lists, and page lists **read** daily `top_queries` / `top_pages` breakdowns for the lookback (so 28-day and 90-day windows differ). Page diagnose, movers, gaps, and questions still read query×page rows (5,000-cell sample ordered by impressions) and can miss tail queries. Treat those as a bounded sample, not every Google row.

## Search questions vs gen-AI exhaust

`GET /analytics/search/questions` (MCP: `get_search_questions`) buckets warehoused query strings: interrogatives, near-me, and **gen-AI exhaust** (queries that look like assistant-style phrasing). Those strings are keywords for IdeaClouds. They are **not** Playbook AI visibility, ranking in generative answers, or a live Google generative report.

## Endpoints

All of these require `analytics:read`. See [`analytics:read`](/scopes/analytics-read) for the full table.

Typical query: `lookback_days` (`28` or `90`). Page performance also requires `page_url`.

## MCP tools

The same facts are MCP tools (`get_search_snapshot`, listings, opportunities, gaps, changes, decay, questions, alerts, create-brief, page performance, traffic, demand). Cards render for snapshot, opportunities, gaps, changes, alerts, create-brief, and page performance. Connect with OAuth and a saved rooftop grant as in [Hrizn MCP](/guides/hrizn-mcp).