Google Search Console API: export more than 1,000 rows

The Google Search Console API is useful when a manual export is too small or too difficult to reproduce. It lets you retrieve search performance programmatically, preserve a request configuration and refresh a report without clicking through the interface. A complete extraction cannot stop at the first response. Search Analytics uses a default rowLimit of 1,000, accepts up to 25,000 rows per response and uses startRow for pagination. Even after paging, Google returns top rows within internal limits, so the result is not a complete event log of every query.

This guide focuses on pages, queries, clicks, impressions, CTR and position. It separates Search Analytics from URL Inspection and sitemap management, uses read-only access and provides a numerical check for the extraction. Store a request manifest beside every output: property, dates, search type, dimensions, filters, data state and retrieval time.

What can the Google Search Console API do?

The Search Console API contains several resources with different purposes. They can read performance data, list accessible properties, manage sitemaps and inspect information about the indexed version of an individual URL. Google's API overview explains that a caller must have the relevant Search Console permission for a property.

This distinction prevents a common category error. Search Analytics is not an indexing test, and URL Inspection is not a bulk live-page checker. The URL Inspection method reference states that the response concerns the version in Google's index.

If you first need to identify which interface report answers each question, read the Search Console capabilities overview. API work should start after choosing the relevant report and aggregation level.

Why does a default export stop at 1,000 rows?

The Search Analytics request body makes rowLimit optional. If you omit it, the service uses 1,000. The accepted range is 1 through 25,000. The startRow field is the zero-based index of the first row in the next slice. These values are documented in the query method reference.

Changing rowLimit to 25,000 increases only one response. To continue, repeat the same request with startRow set to 25,000, then 50,000. Dates, dimensions, filters, search type and data state must remain the same. If any of them changes between pages, the rows are no longer slices of one extraction.

{
  "startDate": "2026-09-01",
  "endDate": "2026-09-01",
  "type": "web",
  "dimensions": ["page", "query"],
  "rowLimit": 25000,
  "startRow": 0,
  "dataState": "final"
}

The property appears in the request URL. A URL-prefix property includes its protocol and trailing slash, such as https://example.com/. A Domain property uses the form sc-domain:example.com. These are syntax examples and do not identify the analyzed project.

How do you export more than 1,000 rows without gaps between batches?

Stop when a response is empty or contains fewer than the requested 25,000 rows. Google's Search Analytics query guide explains an important boundary case: when the available row count is exactly divisible by the batch size, the next request is empty. That empty response proves the end of the slice; it is not a failure.

start_row = 0
batch_size = 25000

while True:
    body["startRow"] = start_row
    body["rowLimit"] = batch_size
    rows = request_search_analytics(body)
    save(rows)
    if len(rows) < batch_size:
        break
    start_row += batch_size

Our calculation. If one date and one search type expose exactly 50,000 rows, the loop makes three calls: two responses with 25,000 rows and one empty response that confirms the boundary. Running the same daily extraction for a 31-day month and one search type can therefore require 93 calls: 31 × 3. This is a request-count calculation for that scenario, not a claim that every property returns 50,000 rows.

Record the offset and row count for every batch. After combining the files, test the uniqueness of the selected dimensions, for example date + page + query + country + device. A duplicate key suggests that a batch was stored twice or that request settings differed. It does not, by itself, demonstrate a Google data error.

Why does pagination still not guarantee every query?

Search Analytics sorts results by clicks and warns that internal limitations restrict the response to top rows rather than guaranteeing every data row. Google's guide to getting performance data states that the method exposes at most 50,000 rows per day and search type. “Full export” therefore means the largest available result for a defined request, not every anonymized or rare combination.

Google recommends requesting data separately by day and search type when the aim is to retrieve more available rows. A wide date range is convenient but does not increase the daily ceiling. Splitting by country or device creates additional requests and may change aggregation. Do that for a stated analysis question, rather than merely to make a file larger.

How do you authorize access without granting unnecessary permissions?

User data requires OAuth 2.0. An ordinary API key does not replace Search Console authorization. Google's authorization guide lists two scopes: webmasters.readonly for read access and webmasters for read/write access. A performance export needs only the read-only scope.

1. Create a project in Google Cloud Console and enable the Search Console API.
2. Configure an OAuth consent screen and create a client for the application type.
3. Request https://www.googleapis.com/auth/webmasters.readonly.
4. Authorize a user who has access to the required property.
5. Treat the refresh token as a secret and keep it out of reports and source control.

Google lists these prerequisites in its application preparation guide. Existing Search Console permissions still apply. A token cannot reveal a property that the user is not allowed to access.

Which quotas matter before a bulk extraction?

Quotas differ by resource. Google's current usage limits allow 1,200 Search Analytics queries per minute for a site and for a user, with separate project limits. Search Analytics also applies load quotas. Grouping or filtering by page and query is more expensive, and longer date ranges consume more load.

URL Inspection has a separate per-property limit of 2,000 inspections per day and 600 per minute. That budget cannot provide performance metrics and should not be spent repeatedly inspecting unchanged URLs. Handle quota errors with delayed retries and increasing intervals instead of a tight infinite loop.

How should you validate the resulting CSV or table?

Compare settings before comparing rows. Two files need the same date interpretation, search type, final or all data state, dimensions, filters and property. Then calculate rows per batch, earliest and latest dates, unique keys, clicks and impressions. Recalculate aggregate CTR as sum of clicks / sum of impressions. Do not average already averaged positions without a defined weighting method.

For a one-off analysis, use the GSC full export tool. It removes the need to configure an OAuth client or write a paging loop, but it cannot remove the source API's limits. Keep the request settings and retrieval date beside the file.

What does the Search Console API not do?

The API does not add a site to Google's index, submit arbitrary pages for accelerated indexing or reveal hidden queries. It is not a fixed-rank tracker: the position field is an aggregate Search Console metric. URL Inspection does not run a live test, and Search Analytics does not explain the cause of an increase or decrease.

Automation also cannot repair an unclear data model. A large file becomes misleading when search types, periods and properties are combined without explicit columns. An automated report is useful only when its request can be repeated and explained.

What should you do after the first extraction?

Save the request manifest and raw response before building derived metrics. Keep daily aggregates separate from dimension-level rows. Then answer one question, such as which pages receive impressions without clicks or which queries changed between comparable periods. The guide to first Google impressions and clicks shows why a short spike should not automatically be treated as a lasting trend.