By Key Takeaways:
- OpenAI launches deep research, a new tool for ChatGPT Pro users that helps with in-depth investigations.
- It gathers information, analyzes data, and compiles reports with sources and explanations.
- The tool can take 5 to 30 minutes to generate results and sometimes makes mistakes or struggles with misinformation.
- AI companies like OpenAI and Google are racing to create smarter AI assistants that can actually help with real-world tasks.
OpenAI just introduced a game-changing research tool for ChatGPT, designed to help users find and analyze information more effectively.
Unlike a simple chatbot response, this feature—called deep research—can actually track down sources, analyze data, and build structured reports in a way that mimics a real research analyst.
An example of ChatGPT’s “deep research” mode showing its progress and citations. Source: OpenAI
Right now, it’s only available to ChatGPT Pro users (who pay $200/month), but OpenAI says it plans to expand access over time.
How It Works
On Humanity’s Last Exam(opens in a new window), a recently released evaluation that tests AI across a broad range of subjects on expert-level questions, the model powering deep research scores a new high at 26.6% accuracy. This test consists of over 3,000 multiple choice and short answer questions across more than 100 subjects from linguistics to rocket science, classics to ecology. Compared to OpenAI o1, the largest gains appeared in chemistry, humanities and social sciences, and mathematics. The model powering deep research showcased a human-like approach by effectively seeking out specialized information when necessary.
Instead of just answering a question instantly, deep research takes a step-by-step approach:
- It searches for data across the internet.
- It backtracks and revises if needed.
- It puts together a final report with citations, summaries, and key insights.
Users can upload PDFs, spreadsheets, or even images to give the AI more context. The entire process takes anywhere from 5 to 30 minutes, depending on the complexity of the request.
For example, if you ask ChatGPT to analyze trends in the retail industry over the last three years, deep research will:
- Pull information from various sources.
- Check for updates or inconsistencies.
- Summarize the key points in a detailed report.
One unique feature? It shows its steps—so you can see exactly how it came to its conclusions, rather than just trusting the output blindly.
What’s the Catch?
Limitations of OpenAI’s Deep Research Model
While deep research sounds impressive, it still has some major limitations:
- It can “hallucinate” facts, meaning it might make things up or misinterpret data.
- It struggles with misinformation, sometimes failing to tell the difference between a credible source and a rumor.
- It takes time—unlike regular ChatGPT responses, you’ll be waiting a bit longer for results.
OpenAI admits these issues and says it’s working to improve accuracy in future updates.
AI Companies Are Racing to Build Smarter Assistants
This launch is part of a bigger trend in AI development. OpenAI and its competitors are shifting away from simple chatbots and focusing on “AI agents” that can actually complete tasks.
Just last week, OpenAI launched Operator, a tool that lets ChatGPT browse the web, book services, and complete online tasks for users. Meanwhile, Google is working on Project Mariner, a similar AI research tool expected later this year.
With AI assistants getting smarter, companies are betting big on features that go beyond chat—moving toward AI that actually helps with real work.
Who Can Use It?
Right now, deep research is only available to ChatGPT Pro subscribers, who get 100 research queries per month. OpenAI says it will gradually expand access to other users, including those on the Plus, Team, and Enterprise plans.
For professionals who spend hours digging through reports, analyzing trends, or fact-checking information, this could be a game-changer—as long as they remember to double-check the AI’s work.
What's your reaction?
