ChatGPT has a powerful feature called Deep Research. It combines two recent developments of ChatGPT, the inference model and the ChatGPT Search function, to enable deep research on any topic you want. This is the kind of feature that has long been lauded as a revolutionary application of Artificial Intelligence (AI) .
OpenAI even called Deep Research "agentic". While AI agents are a hot topic right now, in this case, it simply means ChatGPT can browse the web on its own without any intervention from you. It does have a certain degree of autonomy, but it's still far from Skynet.
Let's explore how Deep Research works in the following article!
How does ChatGPT Deep Research work?
OpenAI's o3, o3-mini, and o1-mini inference models use chain reasoning to break down complex tasks into smaller, distinct steps. This allows them to solve far more complex logic, programming, and scientific problems: Instead of rushing to any seemingly logical answer, they can take the time to consider different options. While errors and erroneous reasoning are still possible, they make far fewer fundamental mistakes and can answer significantly more difficult questions than conventional large language models (LLMs) .
Deep Research uses a version of O3 specifically optimized for research and combines it with ChatGPT's web browsing capabilities. By combining a logical approach and problem analysis with web search capabilities, Deep Research can conduct in-depth research on any topic you choose, then compile a detailed report of its findings. Depending on the difficulty of the question, this process can take anywhere from 5 to 30 minutes. (Deep Research results can be several thousand words long).
To generate these reports, Deep Research is capable of browsing dozens of websites, performing multiple web searches, aggregating statistics from various sources and performing statistical analysis on them, reviewing cited scientific literature, comparing multiple products, organizing data into tables for easy comparison, researching hard-to-find information, and generally conducting in-depth web searches, adjusting its search strategy as needed.
Note : While you can upload files and images to Deep Research for additional context, it doesn't seem to account for them in the research process. And although it will soon have the ability to create charts and visualizations of data.
How to create a ChatGPT Deep Research prompt
Since they weren't professional prompt designers, many people decided to ask Deep Research the best way to set up a prompt for it. This is the template that Deep Research's ChatGPT provided.
Deep Research template
Title : [Brief title and description]
Context : [Provide background information on the topic. Why is this research important? What specific aspect do you want to explore?]
Research focus :
- Main question : [List the specific questions you would like answered]
- Comparison : [If applicable, please mention any comparisons you require]
- Data/Criteria : [Any key factors you wish to consider, such as cost, effectiveness, adoption rate, etc.]
Scope :
- Timeframe : [e.g., past 5 years, future forecast]
- Geographic focus : [e.g., Global, United States only, Europe]
- Industry/Field : [e.g., Finance, Healthcare, Technology]
Preferred format : [e.g., report, table, summary, list of key points]
Preferred sources : [e.g., academic articles, government reports, industry case studies]
Language : [Specify if you wish to study in a specific language]
Instructions for using ChatGPT Deep Research
To use ChatGPT Deep Research, you need to be a subscriber to ChatGPT Plus , ChatGPT Teams, ChatGPT Enterprise, ChatGPT Education, or ChatGPT Pro . ChatGPT Pro subscribers currently receive 120 Deep Research queries per month; other subscribers receive 10 queries.
- Log in to ChatGPT or the ChatGPT app.
- Select Deep Research below the question input field.
- Enter your prompt.
ChatGPT may ask some additional questions, and then conduct the research. Depending on the complexity of the query, it can take anywhere from 5 to 30 minutes to complete.