'Deep Research' From OpenAI Can Do Online Research That Matches Human Analysts, And Faster

OpenAI Deep Research

Conducting thorough research involves several essential steps.

First, a clear research foundation is necessary, outlining the purpose and scope of the investigation. This includes building a solid understanding of the topic through background knowledge and gathering materials from a variety of sources, such as the internet, books, journals, and databases.

Additionally, researchers may need to conduct interviews, make observations, and carry out case studies or surveys to gather primary data.

Throughout the process, critical thinking is crucial to analyze and interpret the findings. Finally, effective writing tools are required to compile the research into a coherent and well-organized report.

OpenAI, the creator of ChatGPT, the Large Language Model-maker that started the generative AI arms race, wants to eliminate pretty much all the above.

That, is by introducing what it calls the 'Deep Research' feature.

In the announcement, OpenAI said that:

"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."

"Deep research is OpenAI's next agent that can do work for you independently—you give it a prompt, and ChatGPT will find, analyze, and synthesize hundreds of online sources to create a comprehensive report at the level of a research analyst."

Leveraging the AI's ability to browse the web, and conduct reasoning based on online searches, Deep Research is able to interpret, and analyze massive amounts of text, images, and PDFs on the internet, pivoting as needed in reaction to information it encounters.

But what makes it unique, is its ability to synthesize knowledge.

With this ability, the Deep Research agent can process information it never heard of, or never been trained on.

This happens because the AI has the ability to synthesize knowledge, which according to OpenAI, is "a prerequisite for creating new knowledge."

To do this, Deep Research can independently discover, reason about, and consolidate insights from across the web.

To power Deep Research, the company refrained from using GPT-4o.

While GPT-4o is ideal for real-time, multimodal conversations, researches require multi-faceted, domain-specific inquiries where depth and detail are critical.

Research also requires extensive exploration and cite each claim.

According to OpenAI, the agent needs to have reasoning capabilities to bridge the gaps between the facts it obtained.

OpenAI trained the AI on real-world tasks requiring browser and Python tool use, using the same reinforcement learning methods behind OpenAI o1, the company's first reasoning model.

While o1 demonstrates impressive capabilities in coding, math, and other technical domains, many real-world challenges demand extensive context and information gathering from diverse online sources.

This is why OpenAI opts to use the reasoning model OpenAI o3 instead, which allows the AI to take on the types of problems people face in work and everyday life.

This way, the agent can use its reasoning ability "to synthesize large amounts of online information and complete multi-step research tasks for you."

"For this reason, deep research marks a significant step toward our broader goal of developing AGI, which we have long envisioned as capable of producing novel scientific research."

Deep Research is designed to assist users engaged in intensive knowledge work across fields such as finance, science, policy, and engineering, where precise, thorough, and reliable research is essential.

Beyond professional applications, it also caters to discerning shoppers seeking hyper-personalized recommendations for purchases that require careful consideration, such as cars, appliances, and furniture.

Each output is fully documented, complete with clear citations and a summary of the reasoning process, ensuring easy verification and reference. It excels at uncovering niche, non-intuitive information that would otherwise require extensive browsing across multiple sources.

By streamlining complex, time-consuming web research into a single query, Deep Research frees up valuable time and enhances efficiency.

To use Deep Research, users can access it through the ChatGPT interface, and select the 'deep research' at the message composer.

Users can then enter their query, and have the AI return thorough and comprehensive results.

Users can even attach files or spreadsheets to add context to their question.

Once it starts running, a sidebar appears with a summary of the steps taken and sources used.

Because Deep Research uses o3's reasoning model, and research does take time, OpenAI said that Deep Research may take anywhere from 5 to 30 minutes to complete its work.

"The final output arrives as a report within the chat – in the next few weeks, we will also be adding embedded images, data visualizations, and other analytic outputs in these reports for additional clarity and context."

The tool is indeed helpful in so many ways.

With its reasoning ability, and the way it lay out its results in a way that is detailed and thorough, OpenAI said that the tool might struggle to distinguish authoritative information from rumors.

And again, because it's an LLM-powered AI, it can also fail to accurately convey when it was uncertain about the information it was delivering.

OpenAI said that Deep Research would be available to anyone who subscribed to ChatGPT Pro, a $200-a-month service that provides access to all of the company’s latest tools. It plans to also offer the tool through its other paid services.

Published: 
03/02/2025