OpenAI has introduced a new artificial intelligence (AI) tool, Deep Research, designed to help users conduct in-depth and complex research using ChatGPT.
OpenAI disclosed in a blog post that Deep Research is designed for situations where users need to analyze multiple sources rather than simply receiving a quick summary.
“Today we’re launching deep research in ChatGPT, a new agentic capability that conducts multi-step research on the internet for complex tasks. It accomplishes in tens of minutes what would take a human many hours,” Open AI stated.
According to OpenAI, the new feature is tailored for individuals engaged in “intensive knowledge work in areas like finance, science, policy, and engineering” and also for consumers making “purchases that typically require careful research, like cars, appliances, and furniture.”
Availability and user access
OpenAI revealed that Deep Research is now available to ChatGPT Pro users, with a limit of 100 queries per month. However, the company stated that support for Plus and Team users will be introduced next, followed by Enterprise clients.
- Regarding accessibility in different regions, OpenAI stated that the launch is geo-targeted, with no release timeline for the U.K., Switzerland, and the European Economic Area.
- The company further disclosed that the query limits for paid users would be “significantly higher soon.”
- To use Deep Research, users will need to select the “deep research” option in the ChatGPT composer and enter their query. Additionally, users have the option to attach files or spreadsheets.
OpenAI stated that the research process takes 5 to 30 minutes, with users receiving a notification upon completion. The feature is currently web-only, but mobile and desktop integration is expected later this month.
Planned enhancements
At launch, Deep Research only generates text-based outputs, but OpenAI announced that it intends to introduce embedded images, data visualizations, and other “analytic” outputs soon.
The company also revealed plans to connect the feature to “more specialized data sources, including subscription-based and internal resources.”
- Given AI’s tendency to generate incorrect or misleading information, OpenAI stated that every Deep Research output will be “fully documented, with clear citations and a summary of [the] thinking, making it easy to reference and verify the information,” they said.
- However, OpenAI acknowledged some limitations, including difficulty verifying authoritative sources, failing to indicate uncertainty, and making formatting errors in reports and citations.
More insights
Deep Research is powered by a specialized version of OpenAI’s o3 “reasoning” AI model, which has been optimized for web browsing and data analysis.
According to OpenAI, the model “leverages reasoning to search, interpret, and analyze massive amounts of text, images, and PDFs on the internet, pivoting as needed in reaction to information it encounters.”
- The model can analyze user-uploaded files, generate and refine graphs using Python, embed images and graphs from websites, and cite specific passages from its sources
- To evaluate Deep Research’s accuracy, OpenAI tested it using Humanity’s Last Exam, an assessment consisting of over 3,000 expert-level questions across multiple academic fields. The o3 model achieved an accuracy rate of 26.6%, outperforming competitors such as Gemini Thinking (6.2%), Grok-2 (3.8%), and OpenAI’s own GPT-4o (3.3%).
