ChatGPT Deep Research now runs on GPT-5.2 with source controls and real-time tracking

OpenAI Powers Deep Research with GPT-5.2 Model

OpenAI has rolled out significant upgrades to Deep Research in ChatGPT, switching to its latest GPT-5.2 model to deliver what the company calls "more accurate" research capabilities than ever before1

1

. The AI-powered research assistant, designed to produce in-depth reports by pulling information from multiple sources, now offers professionals enhanced control over their research workflow. According to OpenAI, the updated GPT-5.2 Instant model provides responses that are "more measured and grounded" in tone, making them contextually appropriate while delivering clearer output and more relevant answers to how-to questions2

2

. These improvements matter because they transform how professionals approach complex research tasks, from analyzing market trends to synthesizing technical documents across various industries.

Users Can Now Specify Research Sources and Add Custom Apps

One of the most requested features has arrived: users can now specify which websites Deep Research will use for each report1

1

. This customizable data integration allows you to add multiple relevant sources for ChatGPT to consider, or select an option to prioritize those sources while allowing further online searches. Previously, the only way to ground Deep Research was by specifying website limiters directly in the prompt itself, but the ability to add this information at a system level delivers more accurate and focused responses2

2

. Beyond websites, you can also add specific apps to your research runs—OpenAI's example demonstrates a plug-in for financial markets from the London Stock Exchange Group (LSEG) for more specialized results1

1

. This professional research tool capability proves particularly valuable for organizations relying on proprietary datasets, ensuring findings are tailored to specific needs rather than general web results.

Full-Screen Report Viewer Displays Citation-Backed Insights

The viewing experience has been completely redesigned with a full-screen report viewer that opens in a document-like interface2

2

. You can now view reports directly in ChatGPT without downloading a separate file to read them1

1

. A table of contents sits on the left-hand side of each report, while the right-hand side shows a list of sources so you can jump to more information. The index is interactive, allowing users to click on any heading or subheading to quickly reach that section2

2

. Once satisfied with the results, you can download files in Markdown, PDF, or Word formats1

1

. This streamlined approach ensures every insight is backed by credible references, addressing the critical need for verifiable research in compliance-heavy industries.

Real-Time Tracking Enables Mid-Research Adjustments

Deep Research tasks can sometimes take up to 30 minutes, and the new real-time tracking feature lets you monitor progress or request edits as reports are happening1

1

. You can ask it to add new sources or queries midway through a report, rather than having to rerun the entire report after completion to introduce extra information. Users can also edit a research plan before it begins, which proves particularly useful given that earlier versions only allowed redirection after research started2

2

. If you wanted to alter details before the plan was created previously, you would have to start all over again. This dynamic flexibility ensures research remains aligned with evolving objectives, with structured overviews generated in as little as 30 minutes3

3

.

Long-Context Reasoning and Enterprise-Grade Security Features

GPT-5.2 brings long-context reasoning capabilities that handle intricate research challenges with ease, allowing professionals to compare findings from multiple scientific studies, identify discrepancies, and provide comprehensive analysis3

3

. For organizations, the update introduces enterprise-grade security with permissions that ensure strict control over data and access3

3

. Administrators can define which tools and sources are accessible during research processes, maintaining compliance standards critical for healthcare, finance, or legal services. The features first rolled out on February 10 for ChatGPT Plus and Pro subscribers, with availability coming to other users in the "coming days," though no clear launch date has been set for free users or those on a ChatGPT Go plan1

1

. Watch for how these capabilities reshape professional workflows, particularly in research-intensive fields where accuracy and verifiable sources determine decision-making quality.