OpenAI has introduced a pioneering feature for ChatGPT, dubbed “deep research,” which aims to streamline how users gather, examine and synthesise online information. The development promises to transform tasks ranging from corporate market analysis to consumer product comparisons by significantly reducing the time needed to pore over vast troves of data.
In essence, deep research converts ChatGPT into an autonomous research assistant capable of parsing hundreds of sources across the internet in a single session. The tool, which operates on a specialised iteration of OpenAI’s forthcoming “o3” model, can reportedly digest text, images and PDF documents, then assemble its findings into a concise report. It is presently available only to ChatGPT Pro subscribers, who pay US$200 per month and can conduct up to 100 queries during that period.
By introducing deep research, OpenAI appears to be addressing a notable gap in the artificial intelligence (AI) market. The system can pause and reflect on its process, identify gaps in evidence and adapt its research path accordingly. This “simulated reasoning” goes beyond the simpler, single-step prompting seen in many AI-driven applications, enabling ChatGPT to refine its answers with each iteration.
The company’s move comes amid growing competition in the AI sector, where rival solutions claim incremental progress in advanced online searches. Industry observers have pointed to releases such as DeepSeek’s R1 model as a spur for OpenAI to expand its range of services and bolster ChatGPT’s autonomous capabilities.
Once a user enters a query, deep research begins by mapping out potential sources, drawing on disparate data streams from across the web. It collects and scrutinises the information, highlighting references in a sidebar that shows each step of the process. This transparency lets users track the flow of ideas and check citations directly. The feature then synthesises the material into a digestible report, with additional insights on its reasoning.
For example, an investment analyst might use deep research to study quarterly earnings figures from numerous corporations, gauge sentiment in financial publications and compile a single document summarising projected sector trends. In scenarios where time and accuracy are vital, the ability to generate a thorough breakdown within minutes could give businesses a competitive edge.
Despite its advanced capabilities, deep research still faces certain hurdles. It can struggle to distinguish authoritative data from rumours or unverified claims, occasionally presenting both with the same level of confidence. Moreover, some early testers have reported that it does not always convey uncertainty accurately — an important caveat when the stakes involve high-risk decisions in areas like healthcare or finance.
In practice, this means users must remain watchful. While deep research can expedite discovery and insight generation, it still benefits from human oversight. Analysts, policymakers and consumers alike should cross-check any critical information and treat the tool’s findings as a starting point rather than the final word.
For now, deep research is restricted to web-based access, with integration into mobile and desktop applications reportedly on the horizon. Plans to broaden availability are also under way: more affordable ChatGPT Plus subscriptions and enterprise-level packages are expected in forthcoming months.
Although the technology requires refinement in source verification, it arrives at a moment when AI tools are being adopted more widely than ever. In industries driven by data and tight deadlines, deep research’s ability to vet large volumes of online material in a matter of minutes could prove a compelling proposition.
It remains to be seen whether the feature will redefine the AI research landscape or simply complement existing offerings. Yet for those already reliant on digital assistance for tasks that once demanded multiple hours of manual study, deep research will likely be a welcome addition — and a sign of AI’s rapid progression from simple chatbot responses to full-fledged research support.