Overview
Purpose
Controls use of crawled content for Gemini models and grounding.
Exact user-agent token
Google-Extended
robots.txt
Allow example
User-agent: Google-Extended Allow: /
Block example
User-agent: Google-Extended Disallow: /
Important limitation
A robots.txt verdict describes the published rule for the tested path. It does not prove crawling, indexing, training, citations or traffic.
Frequently asked question
Does this rule control every product from the same operator?
No. Operators can publish separate user-agent tokens for training, search and user-triggered retrieval. Configure each documented token intentionally.