US generative AI major OpenAI has launched a new agentic AI tool, called deep research, in ChatGPT. 

The new feature is powered by a version of OpenAI’s forthcoming o3 model, designed specifically to enhance web browsing capabilities and streamline data analysis. 

Deep research is targeted at professionals engaged in intensive knowledge work across domains such as finance, science, policy, and engineering. 

The AI agent was developed using end-to-end reinforcement learning, focusing on complex browsing and reasoning tasks across various fields, OpenAI explained.  

In a statement, OpenAI said the deep research agent can operate “independently” when prompted and “ChatGPT will find, analyse, and synthesise hundreds of online sources to create a comprehensive report at the level of a research analyst”. 

“It accomplishes in tens of minutes what would take a human many hours,” OpenAI added.  

However; OpenAI noted that deep research is still in its early phases and comes with certain limitations. 

The tool can “sometimes hallucinate facts in responses or make incorrect inferences” and “it may struggle with distinguishing authoritative information from rumours, and currently shows weakness in confidence calibration, often failing to convey uncertainty accurately” the company said. 

The company is initially rolling out the feature for Pro users, allowing up to 100 queries per month. Access will then expand to Plus and Team users, followed by enterprise customers. 

At present, deep research can retrieve information from the open web and uploaded files.  

Looking ahead, OpenAI aims to integrate it with specialised data sources, including subscription-based and internal resources. 

In January 2025, OpenAI launched the Operator AI agent, which can interact with graphical user interfaces such as buttons, menus, and text fields on a screen.   

OpenAI, in collaboration with SoftBank, is also developing Stargate, a $500bn AI initiative aimed at developing advanced infrastructure.