Deep Research is a sophisticated agent that analyzes and combines information from across the internet to help you with complex, multi-stage research projects. Currently available for Professional tier subscribers, with upcoming access for Plus and Team members.
Deep Research can autonomously conduct comprehensive internet research on complex topics. This feature can compress hours of human research time into minutes.
This new system acts as an independent research assistant – when given a research question, it searches across hundreds of online sources including text, images and PDFs to compile detailed analyst-level reports. It’s built on an advanced version of OpenAI’s upcoming o3 model, specially enhanced for web navigation and data processing. The system uses analytical reasoning to search strategically, understand content across formats, and adapt its research approach based on what it discovers.
OpenAI views this capability to understand and synthesize existing knowledge as a crucial stepping stone toward their ultimate goal of developing artificial general intelligence (AGI) that can generate original scientific insights and discoveries.
Why Did OpenAI build DeepResearch?
This advanced research tool is designed for professionals who require in-depth knowledge gathering in specialized fields such as finance, science, policy analysis, and engineering, where accuracy and thoroughness are essential. It’s also valuable for consumers wanting detailed, personalized product research for major purchases like vehicles, home appliances, or furniture.
The system provides full transparency by documenting all sources and explaining its analytical process, making fact-checking straightforward. It excels at uncovering obscure or unexpected information that would normally require extensive manual searching across multiple websites. By automating and accelerating complex web research tasks through a single query, this tool helps users save significant time they would otherwise spend on comprehensive online research.
This AI research system works autonomously to find, analyze, and combine insights from internet sources. It was developed using reinforcement learning techniques – the same approach used in OpenAI’s o1 model – and trained by performing actual research tasks using web browsers and Python tools. While the o1 model showed strong abilities in technical areas like programming and mathematics, many practical challenges require gathering extensive background information from various online sources. This new research tool extends o1’s analytical capabilities to handle these real-world information gathering needs that people encounter in their professional and personal lives.
How to use Deep Reasearch
To use this research feature in ChatGPT, choose the ‘deep research’ option from the message box and type in your research question – it can handle anything from analyzing streaming service competitors to recommending the ideal bicycle for your commute. You can include relevant files or spreadsheets to provide additional context. As the system works, it displays its research progress and sources in a side panel.
Deep Research will take between 5 and 30 minutes to accomplish the tasks and you will get a notification once it is done. The eventual output is a report in the chat. OpenAI will add other features such as embedded images, data visualization and various analytic outputs in the chat report.