How does deep research by open AI work?

Deep Research by OpenAI is an autonomous research agent built on a specialized version of the o3 model that’s been optimized for web browsing and data analysis. When you provide it with a research query, it first breaks the query into smaller, manageable steps. Then it searches across the internet—accessing a wide variety of sources such as web pages, PDFs, images, and even any files you might attach—to gather all relevant information.^[1][4]

As it collects data, Deep Research uses a chain‐of‐thought reasoning process to interpret, compare, and analyze the retrieved information. In other words, it isn’t just pulling facts; it actively processes and synthesizes the material, adapting its search direction if new insights or discrepancies emerge in real time.^[5][3] This process typically takes anywhere from five to 30 minutes, during which the agent compiles all the findings into a structured report complete with citations for every claim made.^[4]

Additionally, Deep Research can perform tasks like running Python code for data analysis, reading user‑provided documents, and producing tables or summaries as needed. Overall, it serves as a digital research assistant capable of handling complex, multi‑step tasks—tasks that normally would take a human researcher several hours to complete.^[1][2]