OpenAI has unveiled a new ChatGPT feature called “Deep Research,” designed to assist users with complex, in-depth research across various fields, including finance, science, policy, and engineering. Unlike standard chatbot interactions, deep research aims to deliver precise, well-cited results by analyzing multiple sources over an extended search period.

Currently available for ChatGPT Pro users with a limit of 100 queries per month, deep research is expected to roll out to Plus and Team users soon, with Enterprise access following later. However, OpenAI has not provided a timeline for its release in the UK, Switzerland, or the European Economic Area.

To use the feature, users can select "deep research" in the ChatGPT composer, enter a query, and even attach files or spreadsheets. Responses can take anywhere from 5 to 30 minutes, and users receive notifications when their research is complete. OpenAI claims that every output will include full citations and documentation to enhance transparency and verification.

The deep research tool is powered by OpenAI’s latest o3 model, optimized for web browsing and data analysis. This model was trained through reinforcement learning to refine its reasoning skills and improve search accuracy. In benchmarking tests, it outperformed competitors like Gemini Thinking and Grok-2, achieving a 26.6% accuracy score on the difficult Humanity’s Last Exam dataset.

Despite these advancements, OpenAI acknowledges limitations. Deep research can still make errors, misinterpret authoritative sources, and struggle with proper report formatting. Additionally, as AI-generated content becomes more prevalent, users must critically assess the information rather than relying on it without verification.

OpenAI plans to enhance deep research with embedded images, data visualizations, and integration with subscription-based and internal data sources in the future. Whether this feature will rival traditional search engines like Google remains to be seen.