Gemini Deep Research evolves into autonomous AI agent

Summary

Google has rolled out a reimagined version of Gemini Deep Research, turning what was previously a niche report-writing helper into a full-blown autonomous research agent powered by Gemini 3 Pro. The upgraded system performs long-horizon reasoning over huge document sets, iteratively planning searches, filling knowledge gaps, and synthesizing results with a stronger focus on reducing hallucinations. Developers can now tap these capabilities through Google’s new Interactions API, embedding Deep Research-style agents directly into their own applications and orchestrating multiple agents in complex workflows. Google is also seeding the ecosystem with a new DeepSearchQA benchmark to evaluate multi-step web research, and says early adopters are already using the agent in domains like financial services, biotech and drug safety. The launch, which coincides with OpenAI’s GPT-5.2 release, underlines how quickly the frontier AI race is shifting from monolithic chatbots to robust, production-grade agentic systems.