Gemini Deep Research

Gemini Deep Research is the agentic research feature inside Google's Gemini app. Given a question, it builds a multi-point research plan the user can edit, then browses up to hundreds of websites, reasons over what it finds and writes a multi-page report that cites its sources. A report usually takes 5 to 10 minutes, and the user can leave the chat and be notified when it is ready. Google requires only a signed-in Gemini account and an age of 18 or over, so the full workflow can be tried before paying. Best fit: analysts, consultants and researchers who need multi-source synthesis on complex topics in one automated pass.

Google Search is the default source, and users can add their own Gmail, Drive and Google Chat content, uploaded files and Gemini Notebook notebooks, or deselect Search to research only their own material. Finished reports open in Canvas, where they can be exported to Google Docs, shared, or used to create an Audio Overview, quizzes and interactive content. The consumer app publishes no MCP connection of its own. Named gaps remain: no access to paywalled academic databases such as PubMed full text, Scopus or Web of Science, and no Zotero or Mendeley citation management.

Pricing is in USD, read from Google's United States plans page. Deep Research is included on every plan, including the free one, and it draws more of a plan's compute-based usage than ordinary chat. Google AI Plus at $4.99 per month gives 2x the free usage limits and a 128k token context window. Google AI Pro at $19.99 per month gives 4x the free limits, a one million token context window and the Pro model for Deep Research reports, which Plus and free users run on Thinking. Google AI Ultra is sold at $99.99 per month for 5x and $199.99 per month for 20x the Pro limits, and its reports can include charts, diagrams and interactive simulators. Deep Research also has daily and concurrent request caps. Google One offers annual billing as well as monthly.

Developers can run the same agent through the Gemini API. The Gemini Deep Research agent is in preview, is available only through the Interactions API and comes in two versions, Deep Research for speed and Deep Research Max for the most thorough reports. It can search the web, read pages, run code, search uploaded document stores and connect to remote MCP servers for external tools. Pricing is pay-as-you-go on the underlying model and tools, and Google estimates about $1 to $3 for a typical task and $3 to $7 for a Max task at preview rates.

Deep Research is not suited to systematic reviews needing peer-reviewed coverage, because it cannot reach paywalled journals or institutional databases. For academic-database and PRISMA workflows, Elicit ($49/mo billed annually) and Rayyan (free basic tier, paid from $4.99/mo billed annually) provide academic-native pipelines. ChatGPT Deep Research ($20/mo) covers the same need inside ChatGPT, and Perplexity AI ($17/mo billed annually) suits source-display-first workflows. Teams handling confidential material should note three things. Gemini activity auto-deletes after 18 months by default, adjustable to 3 or 36 months or never. Chats already sampled for human review are disconnected from the account and kept up to 3 years. Turning off Keep Activity to stop model training also makes Workspace Connected Apps unavailable.

Google's SOC 2 Type II, ISO/IEC 27001 and ISO/IEC 42001 scope lists name the Gemini app. HIPAA coverage is narrower: it applies to the Gemini app only within Google Workspace under a Business Associate Amendment, so a personal Google AI plan is not covered. Current state Q4 2026: Google AI Pro and Ultra now include Gemini 4 Argon, the free plan runs a Flash-Lite model, Deep Research can draw on Google Chat as well as Gmail and Drive, and the API agent ships as Deep Research and Deep Research Max. Independent review evidence for this listing is the parent Gemini profile on G2 rather than a Deep Research specific corpus.