What is ChatGPT Deep Research?
When people say "openai launches chatgpt deep research," they mean OpenAI has introduced a ChatGPT feature that researches complex questions across multiple sources and returns a documented answer. It can reduce manual research effort for professional analysis, important purchases, and other questions that require source checking, while still requiring users to review citations and conclusions.
If you've ever lost an afternoon to a pile of open tabs, OpenAI has a feature for you. It's called ChatGPT Deep Research.
It takes your question and digs in, pulling from multiple sources to give a comprehensive, sourced response. It suits professionals in STEM, finance and political science, and anyone making an important purchase that needs careful evaluation, such as a car or a home appliance.
How Does ChatGPT Deep Research Work?
Pick the "deep research" option in the ChatGPT composer, type your query and attach any relevant files or spreadsheets. At launch it was available on the web only. The current status section below covers what has changed since.
Once you hit submit, the research takes anywhere from 5 to 30 minutes, depending on how complex your question is. Then you get a detailed, text-only answer with citations throughout. Start it before a meeting and it may well be done by the time you're back at your desk.
Key Features of ChatGPT Deep Research:
- It pulls together information from various perspectives.
- Every response comes with citations, so you can verify the information easily.
- It cuts the hours you'd normally spend on research.
A Reliable and Documented Research Tool
What makes ChatGPT Deep Research stand out is its documentation. Every response is backed by citations, and OpenAI lays out the reasoning behind its conclusions. You can cross-check and verify the data yourself.
AI can still make mistakes, and this tool is no exception. OpenAI says it's working on accuracy, and it uses its specialized o3 "reasoning" model, which processes large amounts of data from multiple sources to produce more precise and reliable results.
The Technology Behind ChatGPT Deep Research
At the core is OpenAI's o3 model, which is built for web browsing and data analysis. It reads complex information in text, images and PDFs. It also uses reinforcement learning, so it gets better over time by learning from experience. (Most of us manage that only some of the time.)
How Does It Benefit You?
- It's designed to pull together reliable information efficiently.
- As the model learns, the quality of its research will keep improving.
Accuracy and Testing
OpenAI tested ChatGPT Deep Research on Humanity's Last Exam (OpenAI's launch post), a tough evaluation with over 3,000 expert-level questions across various fields. The model powering deep research, a version of OpenAI's o3, achieved an accuracy rate of 26.6%. That sounds low, but the exam is intentionally challenging, and other AI models scored even lower. By that comparison, Deep Research already outperforms its competitors.
Limitations to Keep in Mind
No tool is perfect. ChatGPT Deep Research can still get things wrong, especially on ambiguous topics, so read its output the way you'd read a colleague's first draft. It sometimes struggles to seperate credible sources from unreliable ones, and you might run into formatting errors in citations.
The Future of AI Research Tools
OpenAI has big plans for ChatGPT Deep Research. It's looking to add visual elements like graphs and images to go with the text-based responses, which would help, since the answers are text only today.
It's also working on connecting the tool to specialized data sources, including paid and internal ones, for more in-depth research. That connection work has since started to ship, as the current status section below describes. If these upgrades land, ChatGPT Deep Research could become your go-to tool for high-quality research, done in a fraction of the time traditional methods take.
Current status: what has changed since launch
The launch post describes the February 2025 release, and OpenAI has kept updating the product. These are the changes OpenAI lists on its announcement page, in date order.
- February 5, 2025: deep research became available to Pro users in the United Kingdom, Switzerland and the European Economic Area.
- February 25, 2025: all Plus users got access.
- April 24, 2025: OpenAI raised the monthly limits. Plus, Team, Enterprise and Edu users got 25 queries a month and Free users got 5, with a much larger allowance for Pro. A lightweight version powered by a version of o4-mini takes over once you reach the limit for the full version.
- July 17, 2025: deep research gained access to a visual browser as part of ChatGPT agent. You choose "agent mode" in the composer. The original deep research option stays in the tools menu.
- February 10, 2026: you can connect deep research to any MCP server or app and restrict web searches to trusted sites. You can also track progress in real time and interrupt to refine the work with follow-up prompts or new sources.
OpenAI's Help Center describes the current flow. You can start a task with the /Deepresearch command or from the tools menu. ChatGPT proposes a research plan that you can edit before it starts. The result opens as a report with a table of contents and a sources section, and you can download it as Markdown, Word or PDF. It uses connected apps for read actions only. Deep research now runs on OpenAI's latest models by default, and you can choose a legacy model instead. The Help Center no longer lists fixed monthly numbers and says usage varies by plan, so the April 2025 figures above are a dated snapshot. Check your in-product usage counter for your own limit.
Deep research is one of several tools that write cited reports. For a side-by-side look at ChatGPT, Gemini, Perplexity and Claude, plus how to run research on a schedule, read AI agents that research and write reports.
Why You Should Consider Using ChatGPT Deep Research
If you research regularly, for work or personal projects, you know how time-consuming it can be. Picture cutting that time down while still getting reliable, well-sourced information. You start a query, go and do something else for a while, and come back to a cited answer you can check line by line.
Benefits of Using ChatGPT Deep Research:
- It saves time, since you skip the endless scrolling through search results.
- You can focus on the quality of your research rather than the quantity of sources.
- It works for academia, finance or an informed purchase alike.
Conclusion: Embrace the Future of Research
Information overload is the norm, and tool that does the first pass of the reading for you is useful. ChatGPT Deep Research aims to make research easier, faster and more reliable, and the citations give you a way to check that it does. If you've used it, I'd like to hear how it went.
Frequently asked questions
How much does ChatGPT Deep Research cost?
Deep research is part of ChatGPT plans rather than a separate product. OpenAI's Help Center says usage varies by plan and shows a counter of your remaining tasks. In April 2025 OpenAI listed 5 queries a month for Free users and 25 for Plus, Team, Enterprise and Edu users (OpenAI), but check your own counter because those figures can change.
How much effort does ChatGPT Deep Research require?
You select the deep research option in the ChatGPT composer and enter your question. You can attach relevant files or spreadsheets too. Then you wait while the tool reviews sources and produces a cited response.
What risks should users consider before relying on ChatGPT Deep Research?
The tool can make mistakes, particularly when a question is ambiguous or the sources vary in credibility. Inspect the citations and verify important conclusions instead of treating the response as final evidence.
What can go wrong when using ChatGPT Deep Research?
The tool may confuse credible and unreliable sources, and its citations can contain formatting errors. Its responses are text-only in the version this article describes, so graphs and images are not part of that workflow.
What does ChatGPT Deep Research replace?
It cuts the manual work of searching many articles, comparing sources and assembling notes into a documented answer. It still needs your judgment, source review and, for specialized topics, expertise.