What is Google-Extended, and how do I block it?

Google-Extended is Google's robots.txt control token. A control token does not fetch pages. It sets how pages another crawler already collected may be used. From the operator's wording: control token for Gemini training and grounding.

No separate UA string (crawling uses existing Google user agents). The robots.txt token to use is Google-Extended. To tell whether a visit is real, compare the visitor's IP address with the published list.

Add this group to the robots.txt file at the root of your site. It asks Google-Extended to stay out of every page.

To allow it explicitly, for example while a broader rule blocks other bots:

Each Google user agent has its own token, so you can set them one by one. This file blocks Google-Extended and allows Googlebot. Every line comes from the rows in this dataset. Whether allowing a bot keeps you in a product is Google's call, so check each crawler's page.

Google: 'Google-Extended does not impact a site's inclusion in Google Search nor is it used as a ranking signal in Google Search.' Effect on Gemini answers beyond training/grounding use: not stated.

Each other Google user agent has its own robots.txt token, so a rule for Google-Extended does not cover them. See the related crawlers below.

Yes. Robots.txt control token. Google: 'Google-Extended is a standalone product token that web publishers can use to manage whether content Google crawls from their sites may be used for training future generations of Gemini models ... and for grounding' in Gemini Apps and Grounding with Google Search on Vertex AI.

This is a control token, so there is no fetching for it to obey. Googlebot or Applebot reads your robots.txt and applies the token.

Robots.txt is a request, not a lock (RFC 9309).

Google publishes a list at https://developers.google.com/static/crawling/ipranges/common-crawlers.json. We did not download or parse it. A request that claims to be Google-Extended but comes from outside that list is not Google-Extended.