Deep Research

Deep Research is less a single product than a product category that emerged in early 2025: an AI agent that autonomously researches a topic — deciding what to fetch next based on what it finds, looping until it produces a report. ZhenFund's Yusen Dai called these "read-only agents" and noted the category actually started with Google, then OpenAI launched its own Deep Research, Perplexity followed, and Chinese players like Metaso joined with their own versions . In Dai's framing, the PMF was already obvious — "my Deep Research really does a better job than my intern" — because it replicates exactly what an analyst does, with clear willingness to pay among knowledge workers . Elsewhere's coverage treats it as the first wave of agentic AI: as 十字路口Crossing put it when testing Metaso's entry, after ChatGPT and Gemini "fired the first shot," the space became "a fast-iterating, trial-and-error process" where latecomers compete on differentiation like transparent research-reasoning chains and customizable source preferences . OpenAI's version is the reference implementation in most discussions .