Gemini Deep Research

About Gemini Deep Research

Gemini Deep Research is an agentic research feature in Gemini designed for complex research tasks. A user starts by providing a research prompt, which Deep Research transforms into a personalized multi point plan. Gemini presents the plan to the user so it can be reviewed and refined before the research proceeds.

Deep Research can autonomously search and browse up to hundreds of websites while reasoning iteratively over the information it gathers. When selected by the user, it can also draw context from Gmail, Google Drive, and Google Chat. Users can upload their own files to provide additional research material.

During execution, Gemini determines which research subtasks can run simultaneously and which should run sequentially. It continually evaluates gathered information to decide what to investigate next and uses a thinking panel to show what it has learned and what it intends to do next.

Once sufficient information has been collected, Deep Research synthesizes its findings into a structured multi page report. Reports can also be used to create interactive content, quizzes, and Audio Overviews through Canvas.

Key Features

• Research planning: Converts a complex research prompt into a detailed multi point plan that users can review and refine before execution.

• Autonomous web research: Searches and deeply browses the web across up to hundreds of websites to gather relevant and current information.

• Iterative reasoning: Reasons over information as it is gathered and decides what research action to take next.

• Parallel and sequential task execution: Determines which research subtasks can be performed simultaneously and which require sequential execution.

• Workspace research sources: Can use Gmail, Google Drive, and Google Chat as additional research sources when selected by the user.

• Research reporting: Synthesizes gathered information into comprehensive multi page reports that can also be transformed into interactive content and Audio Overviews.