Google has introduced a redesigned version of Gemini Deep Research, positioning it as its most advanced autonomous research agent—one that can generate long-form research reports and also be embedded directly into third-party products via an API.
The announcement landed the same day OpenAI released GPT-5.2, putting the AI research-agent race squarely back in the spotlight.
What’s New: Powered by Gemini 3 Pro + Interactions API
Google says the upgraded Deep Research agent is powered by Gemini 3 Pro and is now available to developers through the Interactions API, which is designed for stateful, long-running, tool-using agent workflows.
The Interactions API is positioned as a unified layer for model + agent interactions, including orchestration and longer tasks—crucial for research agents that don’t finish in one quick response.
Use Cases: From Due Diligence to Scientific Research
Google frames Gemini Deep Research as an agent built to process large information piles end-to-end—highlighting use cases like due diligence and drug toxicology safety research
Google also says it plans to integrate the agent into its own products, including Google Search and NotebookLM, extending “research-agent” behaviors into consumer workflows.
Accuracy, Hallucinations, and Long-Running Tasks
Google explicitly emphasizes that hallucinations become more dangerous in multi-step, long-horizon research, and positions this version as designed to reduce that risk while producing detailed, cited outputs.
DeepSearchQA: Google’s New Open Benchmark for Research Agents
Alongside the agent, Google released DeepSearchQA, a benchmark intended to evaluate agents on complex, multi-step information-seeking tasks, and made it open for broader community use.
Why the Timing Matters: Same-Day Pressure From GPT-5.2
This rollout drew extra attention because OpenAI released GPT-5.2 the same day, describing improvements across long-context understanding and agentic tool use, reinforcing how central “agents that do work” have become in the product race.
Frequently Asked Questions About Gemini Deep Research
What is Gemini Deep Research?
An autonomous research agent that plans and executes multi-step research and produces detailed, cited reports.
What model powers it?
Google says it is powered by Gemini 3 Pro.
How can developers use it?
Through Google’s Interactions API, designed for tool orchestration and stateful, longer-running agent workflows.
What’s the Interactions API?
A unified API interface for interacting with Gemini models and agents, simplifying state and complex interactions; currently in beta.
What kinds of tasks is it aimed at?
Google highlights due diligence and complex scientific research workflows such as toxicology safety research.
What is DeepSearchQA?
An open benchmark for evaluating research agents on difficult multi-step information-seeking tasks.
Is Google integrating it into consumer products too?
Google says it’s planned for products like Search and NotebookLM.
Why was GPT-5.2 mentioned with this news?
Because OpenAI released GPT-5.2 the same day, intensifying direct comparisons in “research agent” capability and positioning.
Hinge Founder Justin McLeod Steps Down to Launch AI-Powered Dating Startup Overtone
