Introduction
Google’s Gemini Deep Research tool can now access users’ Gmail, Drive, and Chat data to provide richer, more “personalized” research results, apparently. The feature, powered by Gemini 2.5 Pro, turns the model into an agent that plans and executes tasks rather than producing instant answers.
Main idea
According to Google’s Dave Citron, users can review and approve research plans before Gemini begins analyzing relevant information from the web and now, from their own Google Workspace files such as Docs, Slides, Sheets, and PDFs.
While similar to tools from OpenAI, Perplexity, and Anthropic’s Claude, this expansion raises familiar privacy concerns. Google says data from connected apps isn’t used to train its AI models. However, its privacy notice admits that human reviewers may see some user data to improve services, warning users not to include confidential material.
Conclusion
The company also cautions that Gemini’s output shouldn’t be relied on for professional, medical, or financial advice. Early reviews of Deep Research are very mixed, some praise its capabilities, while others see it as producing the appearance of research rather than genuine insight.
As ever, the promise of smarter AI comes with a reminder: convenience often costs privacy.
