eBay explicitly bans AI "buy for me" agents in user agreement update

Motivations for the Ban

• LLM “buy for me” agents are seen as likely to cause:
  • Hallucinated or mistaken orders, leading to chargebacks, support load, and disputes.
  • More returns and cancellations when bots misunderstand user preferences or miss “gotchas” like “box only.”
  • Abuse at scale: arbitrage, promotion stacking, triangulated purchases, refund scams, and dropshipping schemes.
• Several comments suggest eBay wants:
  • To be the gatekeeper/paid API for any agents shopping on the platform.
  • To pre-empt big LLMs turning eBay into a commoditized backend data source.
• The clause is viewed as a defensive/legal hedge: it doesn’t need to be perfectly enforced but lets eBay disclaim responsibility and punish abusive integrations.

Enforcement and Detection

• Some argue it’s “impossible to enforce” because agents can drive real browsers and spoof fingerprints or human-like behavior.
• Others note eBay already does aggressive device/browser fingerprinting and behavioral modeling, and can likely detect many automated patterns.
• Ban is seen as mainly useful after problems occur, not as a hard technical barrier.

Bots, Sniping, and Auction Dynamics

• Users question why “buy for me” bots and scrapers are banned while sniping bots remain tolerated.
• Explanations:
  • Scraping and auto-buy can bypass eBay’s interface and discovery; sniping still runs fully on eBay and keeps humans on-site.
• Long subthread debates:
  • How eBay’s proxy bidding works vs sniping.
  • Whether sniping really helps, given second-price mechanics, and how irrational bidders, “nibblers,” and ghost bidding change incentives.
  • Alternative auction designs (time extensions, higher bid increments) and their tradeoffs.

Proposed and Critiqued AI Shopping Use Cases

• Supportive ideas:
  • Agents that monitor for deals with fuzzy requirements (“cheap home server,” specific vintage guitars, car hunting).
  • Assistive agents for disabled users.
  • Arbitrage/mispricing detection and resale workflows.
• Skepticism:
  • Many would never let an AI autonomously spend even a few hundred dollars, especially on used items.
  • High perceived risk of scams, counterfeits, shipping/customs surprises, and nuisance returns.

Business Model, Data, and Platform Incentives

• Strong view that the real target is “laser-focused” agent buying that:
  • Skips browsing, sponsored listings, recommendations, and impulse buys.
  • Turns eBay into a low-margin “dumb pipe” for transactions.
• Ban is framed as:
  • Protecting eBay’s attention/ad-driven funnel and data as a monetizable asset.
  • A precursor to paid, controlled access for agents, rather than outright elimination.

Seller/Buyer Experience and Returns

• Multiple anecdotes of:
  • High effective fees, complex fee structures, and eBay taking a cut of shipping.
  • Painful return scenarios where sellers lose item, shipping both ways, and fees.
  • Perception that eBay increasingly favors power sellers/dropshippers over casual users.
• Some note competition from Facebook Marketplace and local options, though eBay still wins for niche/national-market items.

Policy, Legal, and Meta-AI Discussion

• Quoted clause explicitly bans “robots, scrapers… LLM-driven bots, or any end-to-end flow that attempts to place orders without human review” without permission.
• Debate over whether user agreements “have to be obeyed” vs practical risk of bans or lawsuits.
• Meta thread about AI-written comments on HN itself, with some users calling certain posts “LLM slop” and worrying about a “dead internet” feel.