New ChatGPT agent: research in Deep Research mode

—

© [2025] [𝘒𝘐-𝘎𝘦𝘯.] 𝘔𝘢𝘳𝘤𝘦𝘭 𝘖𝘩𝘳𝘦𝘯𝘴𝘤𝘩𝘢𝘭𝘭. 𝘉𝘢𝘴𝘪𝘦𝘳𝘦𝘯𝘥 𝘢𝘶𝘧 𝘞𝘦𝘳𝘬𝘦𝘯 𝘷𝘰𝘯 𝘈𝘯𝘥𝘳𝘦𝘢𝘴 𝘖𝘩𝘳𝘦𝘯𝘴𝘤𝘩𝘢𝘭𝘭.

The digital business magazine t3n recently analysed the “ChatGPT Deep Research” feature. We have summarised the findings briefly because the tool is specifically designed for research.

Rather than simply producing a list of links, Deep Research generates a fully structured, source-based analytical report and can ask follow-up questions. This makes the mode particularly suitable for complex tasks such as market analyses, competitor comparisons or white papers that might otherwise take hours or even days.

ChatGPT’s standard chat mode is reactive: web search responds to gaps in the chatbot’s knowledge and is designed to provide quick answers. Deep Research takes more time, but works more proactively. It handles queries more thoroughly by breaking them down into several steps and using a dedicated reasoning model.

A clearly formulated prompt remains essential. According to the article, it should include: (1) a concise objective or research question, (2) the target audience and desired tone, (3) the structure in which the information should be presented, and (4) any geographical or time-related limits on sources, or other areas of focus. Users can also upload their own files, such as PDFs, Excel spreadsheets and images, to provide additional context for the research.

With Plus, Team or Enterprise subscriptions, users can make up to ten Deep Research requests per month, followed by an additional 15 requests using a less advanced version. Pro users have access to as many as 125 requests, while free users are limited to five (as of November 2025).

To activate the feature, users need to select “Deep Research” from the drop-down menu in the message field. The interface also shows how many requests remain.

By Maria Kleiner

Source: t3n digital pioneers, 21 November 2025, by Ralf Theil