The name Deep Research is used by several artificial intelligence companies to describe products which combine large language models (LLMs) with internet search capabilities, and generate cited reports on a user-specified topic by letting the LLM autonomously browse the web to find information.[1] Due to their prominence in the field, the name is most associated with OpenAI's product, which is integrated in ChatGPT;[2] however, before its release, Google had already introduced a feature called Deep Research for its Gemini model.[3]
According to OpenAI, OpenAI Deep Research can interpret and analyze text, images and PDFs, and will soon be capable of producing visualizations and embedding images in its reports.[4] However, it can make factual hallucinations or incorrect inferences,[4] can have difficulty distinguishing authoritative sources from rumors,[5] and may not accurately convey uncertainty.[6] It scored 26.6% on the "Humanity's Last Exam" benchmark, surpassing rivals like DeepSeek's model R1 (9.4%) and GPT-4o (3.3%).[7]
As of February 2025, OpenAI's ChatGPT Pro plan (costing $200/month) allows 200 Deep Research queries per month. OpenAI planned to expand this to Plus, Team, and Enterprise users.[6]
In February 2025, Perplexity announced a product with the same name, offered for $20/month.[8]
French
Deutsch.svg){kind=logo}
