Deep Research. ChatGPT Just Got a Huge Upgrade

AI is slowly changing how people do research and OpenAI just took a big step forward with its latest ChatGPT update. The new "Deep Research" tool promises to do the heavy lifting for you—analyzing sources, pulling insights, and delivering detailed reports in a fraction of the time it would take a human.

In a nutshell:

• OpenAI’s new AI tool "Deep Research" is designed for finance, science, policy, and engineering but can also help with buying decisions.
• It pulls from hundreds of sources, analyzes them, and delivers a detailed report in minutes.
• Some users say it’s a major leap for AI, while others highlight issues like hallucinations and trouble with common sense.
• Early tests show it outperforms competitors like DeepSeek R1 and Gemini Deep Research, but still isn’t perfect.
• Available now for ChatGPT Pro users.

So how does it work? You type in a query, select 'deep research' in ChatGPT, and let the AI do its thing. It scans text, images, and PDFs across the web, processes the data, and puts together a well-structured report. OpenAI says it’s built on the upcoming o3 model, optimized for browsing and deep analysis.

A Bold Move Into Knowledge Synthesis

Sam Altman is hyping it up, saying this AI can handle "a small percentage of all economically valuable tasks" and hints that future updates will go beyond analyzing knowledge - they’ll start creating new knowledge too.

Deep Research is aimed at serious fields like finance, science, policy, and engineering. But OpenAI also says it’s useful for product research, meaning it could shake up how people make purchasing decisions - maybe even challenge Google’s role as the go-to for product comparisons.

"Today we are launching our next agent capable of doing work for you independently — Deep Research.

Give it a prompt and ChatGPT will find, analyze & synthesize hundreds of online sources to create a comprehensive report in tens of minutes vs what would take a human many hours. Powered by a version of OpenAI o3 optimized for web browsing and python analysis, deep research uses reasoning to intelligently and extensively browse text, images, and PDFs across the internet." - company said in the tweet.

Real-World Performance is Fast, but Not Perfect

Early users are already testing it, and reactions are mixed. Some are blown away by its speed and depth.

• @levie on X said, "Can confirm OpenAI Deep Research is quite strong. In a few minutes it did what used to take a dozen hours."
• @DeryaTR_ shared a wild story: "I had Deep Research do a patent search & then write a patent for a friend’s cancer-related discovery. It did a phenomenal job & produced a flawless 25-page patent application—that would’ve cost ~$10,000!"

But not everyone is convinced. Some reviewers say it still struggles with hallucinations, misinterprets data, and has trouble distinguishing facts from rumors. 

A recent YouTube review by Ai Explained put it to the test against DeepSeek R1 and Google’s Gemini Deep Research—and the results were mixed.

The Good

• Great for obscure knowledge. It performed well on difficult research tasks, scoring 72-73% on a test where GPT-4 with search only managed 15%.
• Can synthesize massive amounts of data. It pulled relevant information from multiple sources and structured it well.
• Useful for niche research. It outperformed competitors when tasked with digging through specialized data, like analyzing a newsletter’s past ratings.
• Better than DeepSeek R1 and Gemini. It consistently provided deeper insights than these alternatives.

The Bad

• Hallucinates facts. It sometimes made up information or cited sources incorrectly.
• Struggles with common sense. It asked multiple clarifying questions instead of answering simple ones directly.
• Overly aggressive questioning. It often kept asking for more details even when it had enough to proceed.
• Expensive and limited access. Costs $200/month and requires a VPN in Europe.

The Future of AI Research

Despite these flaws, OpenAI expects rapid improvement as more people use the tool.

Deep Research is available now for ChatGPT Pro users. It’s a glimpse into a future where AI handles the tedious parts of research, freeing people up to focus on decision-making.

Will it replace human researchers? Not yet. But as AI keeps improving, it’s clear that knowledge work is changing fast.