New Features Added to Deep Research in ChatGPT

OpenAI has introduced significant upgrades to its Deep Research tool in ChatGPT, enhancing how users conduct detailed, multi-step research. First launched in February 2025, the agentic research assistant was designed to help analyze complex subjects. One year later, the company has refined the experience with additional controls and a new full screen report viewer, offering users greater flexibility in both creating and reviewing research outputs.

One of the major updates focuses on improved research controls. Users can now limit their research to specific websites directly through built-in settings, eliminating the need to manually include such instructions in prompts. This long requested feature enables more accurate and targeted results. Additionally, users can connect trusted apps as sources and edit the research plan before execution something that previously required restarting the entire process.

The tool also retains the ability to adjust research direction mid run, ensuring flexibility as new insights emerge. These enhancements are aimed at making Deep Research more practical for professionals, students, and analysts who require structured and customizable workflows when working with large amounts of information.

On the viewing side, OpenAI has redesigned the interface with a new sidebar entry and a full-screen report viewer. The document style layout includes an interactive index on the left and citations on the right, allowing users to quickly navigate sections and verify sources. Alongside these updates, OpenAI has refined the GPT 5.2 Instant model, making responses more measured, clearer, and better aligned with contextual and advisory queries.