Google-Extended
Google-Extended is a robots.txt control token — not a crawler — that web publishers use to opt content out of being used to train and ground Google's Gemini models and Vertex AI, without affecting Google Search.
Why it matters: it's the only lever that separates 'stay in Google Search' from 'don't be used to train Gemini' — a Disallow on Googlebot would cost you search visibility, but Google-Extended costs you nothing in the SERP.
Google-Extended has no user-agent of its own — crawling is still done by Googlebot’s normal user-agent strings. It is purely a robots.txt product token: adding User-agent: Google-Extended / Disallow: / tells Google that content it already crawled may not be used to train future Gemini models or for grounding in Gemini Apps and Vertex AI. We leave Google-Extended allowed on this site — the same open-robots baseline under which Google indexed 10 of our 16 glossary test pages in a median of 3 days (our index-lag data). Because it doesn’t fetch anything and doesn’t affect Search, it’s the clean way to opt out of AI training while staying fully indexed. See our AI crawlers reference for how this differs from the search-index crawlers. (Source: Google — common crawlers.)