OpenAI has unveiled a new AI-powered tool called Deep Research, designed to assist users in conducting thorough, complex research through its ChatGPT platform. Announced in a blog post on Sunday, the feature targets professionals in fields like finance, science, policy, and engineering, as well as individuals making high stakes purchasing decisions, such as buying cars, appliances, or furniture.
Unlike ChatGPT’s standard quick-answer capabilities, Deep Research is tailored for scenarios requiring meticulous analysis of information from multiple sources. OpenAI describes it as a tool for “intensive knowledge work,” offering detailed, precise, and reliable outputs for users who need more than just a summary.
Availability and Functionality
Initially, Deep Research is being rolled out to ChatGPT Pro users, with a limit of 100 queries per month. Support for ChatGPT Plus and Team subscribers is expected to follow within about a month, with Enterprise users gaining access afterward. OpenAI plans to significantly increase query limits for paid users soon. However, the feature is currently unavailable in the U.K., Switzerland, and the European Economic Area, with no specific timeline provided for its release in those regions.
To use Deep Research, users select the “deep research” option in the ChatGPT composer, enter their query, and can optionally attach files or spreadsheets. The feature is currently web-only, with mobile and desktop app integration slated for later this month. Responses can take anywhere from 5 to 30 minutes, with users receiving a notification once the search is complete.
Capabilities and Future Enhancements
For now, Deep Research outputs are text-only, but OpenAI plans to add embedded images, data visualizations, and other analytic outputs in the near future. The company also aims to integrate “more specialized data sources,” including subscription-based and internal resources, to enhance the tool’s utility.
To ensure accuracy, OpenAI is leveraging a specialized version of its o3 reasoning model, which was trained using reinforcement learning on real-world tasks involving web browsing and Python-based data analysis. This model is optimized to search, interpret, and analyze vast amounts of text, images, and PDFs, while also citing specific sources and generating graphs.
Accuracy and Limitations
Despite its advanced capabilities, Deep Research is not immune to the pitfalls of AI. OpenAI acknowledges that the tool can sometimes produce errors, struggle to distinguish authoritative information from rumors, and fail to convey uncertainty. Formatting issues in reports and citations are also possible.
To address these concerns, OpenAI ensures that every Deep Research output is fully documented, with clear citations and a summary of the AI’s reasoning process. This transparency is intended to make it easier for users to verify the information.
In testing, the o3 model powering Deep Research achieved a 26.6% accuracy rate on Humanity’s Last Exam, a challenging benchmark with over 3,000 expert-level questions. While this may seem low, it outperformed competitors like Gemini Thinking (6.2%), Grok-2 (3.8%), and OpenAI’s own GPT-4o (3.3%).
Competition and Broader Implications
OpenAI’s Deep Research enters a competitive landscape, with Google having announced a similarly named AI feature less than two months ago. The tool’s emphasis on detailed, well-cited outputs could appeal to users wary of overly simplistic or unreliable AI-generated summaries. However, it remains to be seen whether users will rigorously analyze and verify the outputs or simply treat them as polished text to copy-paste.
For students, researchers, and professionals, Deep Research represents a promising step toward more reliable and transparent AI tools. Yet, as with any AI system, its effectiveness will depend on users’ willingness to critically engage with its outputs and cross-check its findings.
Final Word: OpenAI’s Deep Research aims to redefine how we approach complex queries, offering a more thorough and transparent alternative to traditional AI summaries. While its accuracy and reliability are still evolving, the tool’s potential to assist in high-stakes decision-making and knowledge work is undeniable. As AI continues to advance, tools like Deep Research could become indispensable—provided users remain vigilant about verifying their outputs.
