TL;DR
eBay has updated its User Agreement to explicitly ban third-party AI shopping agents, reflecting the rapid emergence of “agentic commerce”—tools designed to browse, compare, and purchase products autonomously on behalf of users.
New Terms Target AI-Powered Shopping
Effective 20 February 2026, eBay’s updated terms specifically prohibit “buy-for-me agents, LLM-driven bots, or any end-to-end flow that attempts to place orders without human review” without platform permission.
While previous terms included general prohibitions on robots and automated data gathering, this marks the first explicit mention of AI agents and LLMs—an acknowledgment that these tools have moved from concept to reality.
The Rise of Agentic Commerce
The phrase “agentic commerce” may sound like marketing jargon, but the tools are already operational. OpenAI added shopping features to ChatGPT Search in April 2025, followed by Instant Checkout in September, enabling direct purchases from Etsy and Shopify merchants within the chat interface.
Notably, eBay CEO Jamie Iannone suggested in November that the company might eventually join OpenAI’s Instant Checkout programme—hinting at a more nuanced relationship with AI commerce than an outright ban might suggest.
Implications for Online Marketplaces
eBay’s terms update reflects the challenge platforms face as AI agents become more sophisticated. The distinction between “with permission” and “without permission” proves crucial: eBay isn’t rejecting AI integration entirely but asserting control over how it happens.
For businesses operating on marketplace platforms, this signals growing friction between AI automation tools and platform policies. Companies building or using AI shopping agents will need to navigate an increasingly complex landscape of platform-specific rules.
Looking Forward
The move likely foreshadows similar policy updates across e-commerce platforms as agentic commerce tools proliferate. Businesses should monitor their platform terms and consider how AI shopping tools fit within permitted use cases.