eBay Blocks Third-Party AI Agents From Browsing Platform
eBay appears to have restricted third-party AI agents from browsing its marketplace, adding another obstacle to the emerging field of agentic commerce. The change became apparent over the weekend when Meta’s Muse agent, which had been used by one user for weeks to browse, curate, and summarize watch listings on eBay and other platforms, could no longer retrieve links, prices, shipping details, or condition notes from eBay listings. In some cases, Muse redirected users to outside websites that mirrored eBay listings instead. The Verge reproduced similar failures using other searches, suggesting the block is not isolated to a single agent or query. The development follows earlier efforts by eBay to limit what AI systems can do on its platform. The company previously updated its user agreement to prohibit tools such as buy-for-me agents and large language model-driven bots that place orders without human review. Until recently, however, some agents could still browse eBay and extract listing information for users. eBay spokesperson Sybille St. Arromand told The Verge that AI is now integrated across the platform for discovery, personalization, and selling, but declined to confirm whether eBay had recently imposed new restrictions on AI agents. She said eBay takes a selective approach to third-party access and has restricted agent access in certain cases, including unauthorized scraping and training on eBay data. The company said it is still discussing potential partnerships and evaluating how third-party agents could help new customers discover its inventory while preserving the trusted end-to-end experience buyers and sellers expect. Meta declined to comment. The episode highlights the broader tension between retailers and AI agents. Shopping is frequently presented as a promising use case for agents that can book travel, find deals, and purchase goods on behalf of users. But many retail and service platforms are blocking external agents or allowing them only limited access. Amazon has blocked external agents from browsing or shopping entirely. Target has restrictions that allow agents to browse in some situations but not others, while Shopify has opened its catalog of about one billion listings so shoppers can configure agents to buy items for them. These varying approaches show that companies are still trying to determine how agentic commerce should work. Retail sites are traditionally built for human users with practical limits on time, attention, and purchasing behavior, not for automated systems that can scan large inventories quickly and potentially strain operations. In October, The Verge found that an agent using Muse could repeatedly add products to carts on the consignment site The RealReal, effectively holding listings in users’ carts and disrupting availability. The RealReal later updated its bot detection and prevention system. At the same time, evidence suggests agentic commerce remains limited in scale. Uber said it fulfilled only four rides through third-party agents in the third quarter and about 500 Uber Eats orders during the week of Halloween. DoorDash reported that only several hundred orders came through third-party agents during a recent quarter, compared with roughly 993 million total orders over the same period. As retailers adjust their platforms, the market is still assessing whether AI agents will become a major channel for commerce or remain a niche tool for specialized use cases.
