Imagine you need to compile a complex report, but you don't have all the information readily available. Or you are looking for trends in the retail sector but don't know where to start. OpenAI now introduces deep research, a new feature in ChatGPT that independently conducts in-depth research and clarifies the steps involved.
How deep research works
Instead of merely generating answers, deep research autonomously carries out multiple steps to gather relevant data. The AI processes text, images, and files such as PDFs and spreadsheets to formulate a well-founded response. This process can take 5 to 30 minutes, depending on the complexity of the question. The result is displayed in the chat, including a sidebar with source references and a summary of the research method.
Limitations and Future Developments
Although deep research should function at the level of a research analyst, OpenAI acknowledges some limitations. The AI can sometimes make mistakes, struggle to distinguish reliable sources, and not always accurately assess uncertainty in its answers. However, OpenAI continues to work on improvements, such as the integration of embedded images and graphs in future versions.
What does this mean for companies and professionals?
For companies and professionals, deep research offers an efficient way to obtain comprehensive analyses without spending hours manually searching for information. This can be valuable for market research, trend analysis, and strategic decision-making. OpenAI positions the feature as a premium tool, available to Pro users with a limit of 100 searches per month. Other paid users will have limited access, with promises of faster and cheaper versions in the future.
AI research at a new level
Deep research scores high on accuracy in AI benchmarks. The model achieved a score of 26.6% in a test with expert questions, well above previous versions of ChatGPT. This underscores the growing potential of AI in conducting complex analyses and in-depth research.