OpenAI upgraded ChatGPT's Deep Research feature with GPT-5.2, adding source selection controls and a built-in document viewer. The update rolled out to Plus and Pro subscribers on February 10, with broader availability expected in the coming days.
Deep Research now runs on OpenAI's latest GPT-5.2 model, which the company says makes reports "more accurate" than previous versions. The feature autonomously browses the web for 5 to 30 minutes to compile detailed research reports with citations.
Users can now specify which websites Deep Research uses for each report. The new controls let researchers add multiple relevant sources for ChatGPT to consider, or prioritize selected sources while allowing additional online searches.
Supported apps can also connect directly within ChatGPT, with OpenAI demonstrating a London Stock Exchange Group plugin for financial market research.
A built-in full-screen viewer displays reports directly in ChatGPT without downloading separate files. The interface shows a table of contents on the left and source citations on the right, enabling navigation through lengthy documents. Completed reports export as Markdown, PDF, or Word files.
Real-time monitoring addresses previous limitations where users couldn't modify research once started. Deep Research tasks that previously took up to 30 minutes can now be tracked live, with options to add new sources or queries mid-process rather than restarting entire reports.
The update follows OpenAI's introduction of Deep Research last year, initially powered by the o3 model optimized for web browsing and data analysis. A lightweight version using o4-mini arrived in April 2025 for free users, while paid tiers received higher query limits.
ChatGPT Pro subscribers get 250 queries per month, Plus, Team, Enterprise, and Edu users receive 25 queries per month, and free users get 5 lightweight queries monthly.
ChatGPT Plus and Pro subscribers gained immediate access to the new features on February 10. OpenAI says the same capabilities will reach other users in the "coming days," though no specific timeline exists for free users or ChatGPT Go plan subscribers.
Deep Research represents OpenAI's response to criticism about AI-generated citations and research accuracy. The feature emphasizes authenticated sources and industry-standard data, addressing concerns about unreliable information found through general web searches.
The GPT-5.2 upgrade follows OpenAI's pattern of model improvements for specialized tasks. Earlier versions of Deep Research scored 26.6% on the "Humanity's Last Exam" benchmark, outperforming rivals like DeepSeek's R1 (9.4%) and GPT-4o (3.3%).
OpenAI's documentation shows Deep Research works with authenticated sources, approved URLs, and connected apps to produce "more credible, actionable reports." The system prioritizes recent statistics from regulators, official agencies, and central bank publications, supplementing them with industry benchmarking and strategy research.
Legal professionals have already adopted Deep Research for casework, comparing it favorably against traditional platforms like Lexis and Westlaw. One attorney reported using voice-to-text input for research questions, a feature absent from conventional legal databases, while noting ChatGPT's $200 monthly enterprise access compares to Lexis's $270 subscription.
The update arrives as OpenAI faces ongoing scrutiny about AI research capabilities. A 2023 Duke University study highlighted ChatGPT's tendency to "fabricate or hallucinate citations", noting users searching for AI-generated sources through Google or library databases "will turn up NOTHING."
"OpenAI says Deep Research occasionally makes factual errors or incorrect inferences,"
acknowledging limitations while positioning the upgraded feature as a step toward more reliable AI-assisted research. The improvements come as ChatGPT expands into government and enterprise applications and faces competition from other AI assistants.
