Just six months after its release, ChatGPT Agent, a product of OpenAI, is facing the risk of being shut down. According to the latest reports, the number of weekly active paying users has plummeted from 4 million at its peak to less than 1 million, with a loss rate as high as 75%.

Although it initially attracted about 11% of subscribers who tried it, ChatGPT Agent quickly hit a growth stall. The main reason was that most users could not understand the actual purpose of this tool, and many didn't even know it existed. In addition, slow system performance, poor reliability, and potential cybersecurity risks further pushed users away.

The name "ChatGPT Agent" itself is considered misleading. In fact, other modes of ChatGPT (such as the reasoning model) already have "agent-like" features, such as writing code, web browsing, and image analysis. The only differentiating feature of this product is the "virtual browser," and giving it the name "Agent" led users to mistakenly believe that other functions lacked agent capabilities.

Facing the failure of a general-purpose agent, OpenAI is adjusting its strategy and shifting towards developing more specialized agents with clear functions. For example, the newly launched "Shopping Research" agent is specifically designed for product recommendations and searches.