Google-Extended is the user-agent Google uses to crawl content for Gemini training and other non-search AI features. It is distinct from Googlebot, which crawls for the Google search index (the same index AI Overviews uses).
Blocking Google-Extended in robots.txt opts your content out of Gemini training without affecting Google Search ranking, AI Overviews citation candidacy, or any other classic Google product. This makes Google-Extended unique among AI bots: it is the only one where blocking has no current-citation downside.
Most operators allow Google-Extended for the same reason they allow GPTBot — contributing to model training carries no visible cost and may benefit future-model brand recognition. Operators who specifically do not want their content used for training can block Google-Extended without consequence to Google Search performance.
The user-agent string is: Mozilla/5.0 AppleWebKit/537.36 (KHTML, like Gecko; compatible; Google-Extended/1.0; +https://developers.google.com/search/blog/2023/09/google-extended).