Gemini Deep Research

Gemini Deep Research is a user-started research workflow in Google's Gemini apps. A user gives it a topic, reviews or edits the proposed research plan, and then asks Gemini to produce a report. Google describes the work as in-depth, real-time research rather than a background crawl of the entire web.

Google Search is included as a source by default. Users can change the source selection and may add material such as uploaded files or NotebookLM notebooks. Google says a report usually takes five to ten minutes because Gemini analyzes many sources, with more complex work sometimes taking longer.

There is no dedicated HTTP user-agent token for Gemini Deep Research in this bot. No example user agent, IP range, Cloudflare entry, or Web Bot Auth directory is attached to it either. A generic Google request in a server log should not be labeled as a Deep Research visit without another verified signal.

For public web material, discoverability usually begins with Google Search because Search is the default source. Pages that Googlebot cannot crawl may be absent from that source set. This is an indirect relationship: a Googlebot crawl is not proof that Deep Research selected the page, and a Deep Research report is not a separate public search ranking.

The record does not identify Gemini Deep Research as a model-training crawler or state that pages considered for a report are added to training data. Google provides Google-Extended as the separate robots.txt product token for certain Gemini training and grounding uses. That control does not affect a site's inclusion or ranking in Google Search.

This page marks the workflow as respecting robots.txt, but there is no standalone Deep Research token to target. In practice, publishers should apply the documented control for the underlying use: Googlebot for Search crawling and Search display controls, or Google-Extended for the Gemini uses covered by that token. A made-up GeminiDeepResearch group has no verified effect.