OpenAI has launched a new AI tool, named Deep research, designed to conduct time-consuming online research for users on various topics, from complex scientific inquiries to product recommendations. This tool will be accessible to specific paying customers via OpenAI’s ChatGPT chatbot, as announced in a blog post on Sunday.
What is deep research by OpenAI?
Deep research is positioned as an AI agent, functioning similarly to a research analyst. It has the ability to scour online content such as text, images, PDFs, and user-uploaded files, generating in-depth reports on complex queries. OpenAI claims that it can accomplish tasks typically requiring several hours in a mere “tens of minutes,” with timeframes varying between five to thirty minutes based on task complexity.
The tool offers capabilities that extend beyond simple question-and-answer functionality typical of chatbots. For instance, during a demonstration, Kevin Weil, OpenAI’s chief product officer, showcased how Deep research efficiently compiled a detailed report on Albert Einstein for hypothetical Senate staff preparing for his congressional nomination, producing relevant questions for Senate inquiries.
Deep research run queries iteratively, enabling recursive learning from related searches. The reports generated include citations to the sources of information; however, OpenAI cautioned that the tool could produce fabricated information, a phenomenon referred to as “hallucination.” It may also struggle to differentiate between rumor and authoritative information.
Deep research marks the second AI agent release from OpenAI in 2023, following the introduction of Operator, another tool that assists users with tasks such as booking flights and grocery deliveries. Both AI agents are initially available to subscribers of the $200-a-month ChatGPT Pro model.
OpenAI’s CEO, Sam Altman, has stated that AI agents represent a significant breakthrough for the field, as competition intensifies with emerging technologies from companies such as the Chinese firm DeepSeek.
Deep research has been optimized for extensive web browsing and data analysis, utilizing OpenAI’s new reasoning technology, referred to as OpenAI o3. It is particularly aimed at professionals in fields such as finance, science, policy, and engineering, as well as consumers seeking detailed product recommendations.
At launch, users are limited to 100 queries per month due to the tool’s high computational demands. OpenAI has plans to expand access to its Plus, Team, and Enterprise subscriptions in the future but has not specified a timeline for this rollout.
The tool performed favorably in evaluations of expert-level tasks, automating extensive manual research and scoring 26.6% accuracy on a broad evaluation known as Humanity’s Last Exam, where it outperformed previous models in several domains, indicating significant improvements in information retrieval capabilities.
While the tool boasts advanced functionalities, OpenAI noted that it is still in the early stages and that improvements are expected over time. Users can anticipate enhancements such as increased rate limits and additional features, including embedded images and visual data representations, in forthcoming updates.
Featured image credit: Jonathan Kemper/Unsplash
