OpenAI Launches Deep Research, a New AI Tool for Data Analysis – How to Use It

OpenAI has introduced Deep Research, a new feature in ChatGPT designed to handle complex research tasks that typically take hours. The company claims this tool can complete multi-step research in just minutes, making it particularly useful for professionals in finance, science, policy, and engineering who need structured, in-depth reports.

This new capability allows ChatGPT to search, analyze, and summarize large amounts of information across the web. OpenAI sees this as an important step toward developing more advanced artificial intelligence, capable of producing expert-level insights across multiple fields.

How Deep Research Works

Unlike the standard ChatGPT experience, which focuses on real-time conversations, Deep Research is built for autonomous, structured analysis. It runs on a version of OpenAI's upcoming o3 model, optimized for web browsing, data analysis, and reasoning.

The tool is designed to handle multi-step research workflows by searching across hundreds of online sources, adapting its approach based on new information, and structuring its findings into comprehensive reports. Users can also upload documents, spreadsheets, and other files, which the AI can analyze and incorporate into its research.

To maintain transparency, Deep Research cites specific sentences or passages from the sources it uses, allowing users to verify its findings. It can also generate graphs, embed images from websites, and summarize key points, making its reports more useful for professional decision-making.

Who Can Benefit from Deep Research?

OpenAI has positioned Deep Research as a tool for professionals who need detailed, evidence-based analysis without spending hours digging through sources. Researchers, policymakers, and analysts can use it to streamline their work, while businesses may find it useful for competitive analysis, market research, and strategy planning.

The feature can also serve consumers looking for in-depth recommendations, such as comparing different car models, appliances, or financial products. By analyzing multiple sources, Deep Research can provide more nuanced insights than a traditional search engine.

Performance and Accuracy

Deep Research has been tested on Humanity's Last Exam, a benchmark designed to measure AI performance across over 3,000 expert-level questions in more than 100 subjects. The model achieved 26.6% accuracy, outperforming OpenAI's previous models, including GPT-4o (3.3%) and o3-mini (13%).

It also performed well on the GAIA public leaderboard, which evaluates AI in real-world problem-solving, reasoning, and web browsing. OpenAI claims this marks a significant improvement in AI's ability to automate complex research tasks across various industries.

How to Use Deep Research

Deep Research is currently available through ChatGPT's web version and will soon be rolled out to mobile and desktop apps. OpenAI plans to expand access to Plus and Team users, followed by Enterprise customers.

Using Deep Research is simple-users just need to select the "Deep Research" option in ChatGPT, enter their query, and wait while the AI processes the request. The system typically takes 5 to 30 minutes to complete research, allowing users to continue other tasks in the meantime.

The results include detailed reports with citations, summaries, and visual data, and OpenAI is planning to introduce embedded images and data visualizations in future updates.

How It Compares to GPT-4

While GPT-4 is built for real-time conversations and multimodal interactions, Deep Research is designed for deep, structured analysis. It goes beyond simply answering questions by collecting, synthesizing, and interpreting information from multiple sources.

Limitations and Challenges

Despite its advanced capabilities, Deep Research has limitations. Like all AI models, it may sometimes misinterpret data, hallucinate facts, or struggle to distinguish between authoritative and unreliable sources. OpenAI acknowledges that while it performs better than previous models, it still has room for improvement in fact-checking and handling uncertainty.

Formatting inconsistencies in reports and citations are another minor issue, which OpenAI plans to refine over time. Additionally, the compute-intensive nature of Deep Research means access is currently limited to Pro users, with a cap of 100 queries per month.

Source