Use ChatGPT Agent to find a scarce Labubu toy

Amanda Caswell tested ChatGPT Agent by looking for a Big Into Energy Labubu, a toy she describes as difficult to find. She had already tried Google Search and wanted to compare the effort required when an agent handled the browsing.

Her initial request asked it to find the toy nearby. The agent asked whether she wanted information or a purchase; she chose purchase, setting a concrete endpoint for the task. Caswell watched it inspect multiple websites, read pages and work through the search in its browser view.

She reports that within six minutes the agent found the requested toy, added it to a cart and reached checkout. It then asked for her shipping address and card information. Caswell took control of the agent's browser to finalize the sale, so the reported result includes an explicit human checkout handoff rather than an unattended purchase.

That boundary also appeared in another shopping test in the same article: a DoorDash lunch request needed her ZIP code and handed back payment and address entry. The useful pattern is specific search or selection followed by a visible commitment step, with location and purchase intent supplied when the agent asks.

Prompt
Original prompt

Source: www.tomsguide.com · Amanda Caswell