OpenAI has enhanced ChatGPT’s Deep Research tool with new updates. The AI-powered research assistant originally launched in February 2025 to help users handle complex, multi-step research tasks more easily.
Nearly a year later, the company has introduced advanced control options and a new full-screen report viewer, making the tool more flexible and efficient to use.
Alongside these improvements, OpenAI has also upgraded the GPT-5.2 Instant model to deliver more balanced and context-aware responses.
Greater Control and Improved Accuracy in Research
The most significant highlight of this update is the increased control users now have over the research process. Through ChatGPT settings, users can directly limit research to specific websites.
Previously, users had to manually include website restrictions within the prompt. Now, this feature is available at the system level, which helps generate more accurate and focused results.
This change makes it easier to create reliable research outputs based on preferred sources.
In addition, users can activate connected apps during research. These integrations act as trusted data sources and help generate more detailed and context-rich reports.
Edit Research Plans Before Starting
OpenAI has also introduced a useful feature that allows users to review and edit a research plan before it begins.
Earlier, users could only modify the direction of a project after the research had started. If they wanted to change the scope beforehand, they had to restart the entire process.
Now, users can refine the plan in advance, which saves time and improves workflow efficiency. However, the option to adjust the direction mid-process is still available.
New Interface and Full-Screen Report View
The update also improves the overall user interface. Deep Research now includes a more intuitive sidebar entry point.
The most important addition is the new full-screen report viewer, which provides a document-like reading experience.
It features an interactive index panel on the left and a citations panel on the right.
Users can click on any heading or subheading in the index to jump directly to that section. This makes navigating and understanding long research reports much easier than before.
Improvements to GPT-5.2 Instant Model
Beyond Deep Research, OpenAI has updated the GPT-5.2 Instant model in both ChatGPT and its API.
According to the company, the revised model now delivers more balanced and fact-based responses. Its tone aligns more naturally with the context of the conversation.
The update especially improves clarity and relevance for “how-to” questions, advice, and guidance-related content.
Users can now expect more structured and accurate outputs overall.