Google launches Deep Research and Deep Research Max AI agents that can search web and private data

Google has launched two new AI research agents that can search both the open web and private enterprise data simultaneously. Deep Research and Deep Research Max represent the company’s most significant upgrade to autonomous research capabilities since the product’s debut, offering developers the ability to fuse public and proprietary information through a single API call.

The release comes as tech giants race to build AI systems that can handle the exhaustive, multi-source research that typically consumes hours of human analyst time. For Google, this marks a clear attempt to position its AI infrastructure as the backbone for enterprise research workflows in finance, life sciences, and market intelligence – industries where accuracy is critical.

Two-tiered approach balances speed against thoroughness

Google built two separate research agents instead of one to address a fundamental tension in AI design: the tradeoff between speed and depth. Each agent serves different use cases:

• Deep Research: Optimized for low-latency, interactive applications. Ideal for embedding research capabilities directly into user-facing interfaces like financial dashboards that need to answer complex questions in near-real time
• Deep Research Max: Uses extended test-time compute to spend more processing cycles reasoning, searching, and refining outputs. Designed for background workflows where analysts need exhaustive, fully sourced reports

“Use Deep Research when you want speed and efficiency, and use Max when you want the highest quality context gathering & synthesis using extended test-time compute,” Google CEO Sundar Pichai wrote on X. The Max version achieved 93.3% on DeepSearchQA and 54.6% on HLE benchmarks.

Private data access transforms enterprise research capabilities

The most significant new feature is Model Context Protocol (MCP) support, which allows Deep Research to query private databases, internal document repositories, and specialized third-party data services without sensitive information leaving its source environment.

This means organizations can now point the research agents at their internal databases and external data sources simultaneously. A hedge fund, for example, could have Deep Research analyze its internal deal-flow database alongside financial data terminals and publicly available web information in a single query.

Google is actively working with major financial data providers on MCP integration:

• FactSet
• S&P
• PitchBook

The system accepts multiple input formats including PDFs, CSVs, images, audio, and video. Developers can run searches across Google Search, remote MCP servers, URL Context, Code Execution, and File Search simultaneously – or disable web access entirely to search only custom data.

Native visualizations eliminate manual chart creation

Both agents now generate charts and infographics directly within their reports, addressing a major limitation of previous text-only outputs. This eliminates the friction of having to export data and build visualizations separately.

The charts render as actual HTML or Google’s Nano Banana format within the markdown output – not screenshots or suggestions to visualize data. For enterprise users in finance and consulting who need stakeholder-ready deliverables, this transforms the agents from research accelerators into tools that can produce near-final analytical products.

Additional new features include:

• Collaborative planning that lets users review and refine the agent’s research plan before execution
• Real-time streaming of intermediate reasoning steps
• Granular control over investigation scope

From consumer feature to enterprise infrastructure

Google’s journey with Deep Research has been remarkably rapid. The company first introduced it as a consumer feature in the Gemini app in December 2024. By March 2025, Google upgraded it with more advanced models. The December 2025 release marked the pivot to developer access through the Interactions API.

Today’s release runs on Gemini 3.1 Pro, which Google released in February 2026. That model scored 77.1% on ARC-AGI-2 – more than double the performance of its predecessor on novel logic pattern solving.

Google emphasizes that developers accessing Deep Research through the API tap into “the same autonomous research infrastructure that powers research capabilities within some of Google’s most popular products like Gemini App, NotebookLM, Google Search and Google Finance.”

Competitive landscape intensifies around autonomous research

Google faces growing competition in autonomous research agents. OpenAI is developing agent capabilities within ChatGPT under the codename Hermes, while Perplexity has built its entire business around AI-powered research. A growing ecosystem of startups is also targeting various aspects of automated research workflows.

Google’s advantage lies in combining its search infrastructure – providing access to the broadest web information index available – with MCP-based connectivity to enterprise data sources. No other company currently offers a research agent that can simultaneously query the web at Google Search’s scale and access proprietary data repositories through a standardized protocol.

However, the new agents are only available through the API, not in the consumer Gemini app, leading to some user complaints about Google prioritizing developers over consumer subscribers.

Impact on knowledge-intensive industries

The implications are most immediate for industries that depend on exhaustive research as a core business function. In financial services, where analysts routinely spend hours assembling due diligence reports from scattered sources, Deep Research Max could automate the initial research phase entirely.

In life sciences, Google collaborated with Axiom Bio and found that Deep Research enabled new levels of research depth across biomedical literature. For market research and consulting, the ability to produce stakeholder-ready reports with embedded visualizations could compress project timelines from days to hours.

The critical question remains whether automated outputs will meet the quality and reliability standards that professionals in these fields require. While Google’s benchmark numbers are impressive, real-world research often requires judgment that remains difficult to automate.

Both Deep Research and Deep Research Max are available now in public preview via paid tiers of the Gemini API, with Google Cloud availability for startups and enterprises coming soon.