What Amazon and eBay are actually doing
Amazon has
updated its legal policies
to control exactly how AI agents can interact with its platform, banning behaviors that mimic humans or bypass CAPTCHAs, and reserving the right to cut off access "at its sole discretion." It's also
blocked major AI crawlers
from Meta, Google, and Huawei to prevent them from scraping product pages and prices.
But Amazon isn't just playing defense. It's building its own agent,
Buy with Me
, designed to purchase products even from outside Amazon, while keeping users inside its ecosystem. The message is clear: we'll do agentic commerce, but the agent will be ours.
eBay took a different approach,
banning unauthorized AI agents
from placing orders while piloting its own agent and staying open to partnerships with third parties, on eBay's terms. Think authentication requirements, usage limits, and strict data governance rather than open scraping.
| Marketplace | Strategic Approach | Key Actions & Policies | Main Objectives |
|---|---|---|---|
| Amazon | Defensive + Proprietary Agent Control external access while building own agent ecosystem | • Updated legal policies to ban human-mimicking agents • Blocked AI crawlers (Meta, Google, Huawei) via robots.txt • Developing "Buy with Me" cross-platform agent • Reserved right to restrict agent access at sole discretion | Protect $50B+ annual ad revenue Prevent becoming invisible backend Own the agent interface |
| eBay | Selective Partnership Ban unauthorized agents, enable certified ones on eBay's terms | • Banned unauthorized AI agents from placing orders (Feb 2025) • Piloting own AI shopping agent • Open to third-party partnerships with authentication • Emphasis on governed data access and compliance | Maintain customer relationship Control data exposure Regulatory compliance |
| Shopify* | Protocol-First Building infrastructure for certified agent integration | • Developing agentic commerce infrastructure • Requires human review for "buy-for-me" agents • Creating official checkout integrations • Enabling cross-merchant cart building via API | Enable merchant access to agent economy Standardize integration Maintain transaction control |
| Walmart | Embracing Hybrid Model Build internal agents + accept external traffic | • Accelerating integrated AI assistant development • Accepting referral traffic from ChatGPT • Developing concierge-style shopping experiences • Treating agents as new customer acquisition channel | Capture new traffic sources Modernize shopping experience Compete on convenience |
*Shopify is not a marketplace per se, but operates as one at scale through its merchant network.
Why marketplaces are worried
The biggest fear is simple: if an AI agent can compare offers across platforms and complete purchases without users ever visiting Amazon or eBay, those
browsing sessions disappear
. And the entire retail media model - where brands pay for sponsored placements and search results - depends on those sessions.
AI agents threaten to short-circuit this
by selecting products based on user preferences and value rather than who paid the most for visibility.
There's also a deeper issue:
customer ownership
. In an agent-driven world, shoppers entrust their preferences, payment info, and purchase history to the AI assistant, not the marketplace. The agent becomes the interface; the marketplace becomes invisible infrastructure. For companies that have spent years building loyalty programs and ecosystems, that's a hard pill to swallow.
Add
compliance risks
(agents making mistakes with prices or policies),
reputational concerns
(being blamed for AI hallucinations), and
technical burdens
(poorly designed agents hammering servers), and you have a perfect storm of defensive moves.
The shift from scraping to protocols
Here's where it gets interesting. Instead of blocking agents entirely, the industry is moving toward
agentic commerce protocols, standardized ways for agents to access catalogs, build carts, and complete purchases. OpenAI announced one with Stripe. Shopify is building its own. And Google just launched the
Universal Commerce Protocol. Marketplaces aren't banning agents in this model. They're choosing which ones to work with, negotiating what data gets exposed, and maintaining some control over product presentation. Meanwhile, they're building their own agents to keep the orchestration value in-house. The shift is from "rogue agents scraping everything" to "certified agents with API access," where the power still sits with the platforms.
What this means for merchants
For brands selling on marketplaces, the question isn't who wins this battle. It's how to adapt when your buyer is increasingly an AI rather than a human.
- Data quality becomes critical: AI agents need structured product information: clear titles, complete attributes, consistent pricing, detailed logistics. Listings with messy data won't get recommended, even if they perform well in ads or search. This makes centralized feed management and data discipline, the core of what platforms like Lengow do.
- Think cross-platform: When an agent compares Amazon, eBay, and your D2C site simultaneously, you can't optimize in silos anymore. Price consistency across channels, availability, and delivery promises all matter. Being selected as the "best option" in a multi-platform comparison is the new game.
- Operational excellence matters more: Agents arbitrate on explicit criteria - budget, timing, preferences - not just paid placement. This could level the playing field for sellers who nail the fundamentals but haven't mastered retail media bidding wars.
Marketplaces aren't blocking AI agents. They're trying to control them, steering agentic commerce into walled gardens where they still own the customer relationship and ad inventory.
For brands,
this new reality
means structured data isn't just best practice anymore, it's basically survival. Cross-platform consistency matters more than ever. And the relationship between marketplaces and merchants is being renegotiated in real time.
Adrian Gmelch
Adrian Gmelch is Director of Content at Lengow, where he leads content strategy while staying firmly hands-on: reading the research, and tracking the trends that matter before they go mainstream.
Table of content
Is your catalogue ready for AI Shopping?
Discover what AI assistants look for, how products get selected and what you can do now to stay visible in the new shopping journey.