Amazon is drawing a hard boundary in agentic commerce. On September 21, users on X began posting screenshots of a popup blocking Meta’s Muse assistant from completing purchases on Amazon.com. The popup told users that “continued access by an unauthorized AI agent violates Amazon’s Conditions of Use.”
An Amazon spokesperson told Adweek the company was not informed in advance that Muse would access the store, and it did not authorize the integration. The retailer has asked Meta to remove Amazon from the experience.
Why Amazon is cutting off Muse
Amazon’s argument goes beyond permission. Its spokesperson said third-party shopping agents should operate openly and respect a service provider’s decision to participate, the same way food delivery apps, delivery services and online travel agencies do. For Amazon, Muse raises a few specific problems:
- Account transparency: Amazon says Muse does not identify itself as an agent when accessing customer accounts and processing transactions.
- Data handling: The company claims Muse captures and stores customer credentials and scrapes account data, creating security and privacy questions.
- Experience control: When an agent shops through Amazon, customers bypass personalization, browsing history, recommendations and discovery features Amazon uses to guide purchases.
Those are operational and trust issues, not just technical friction. Amazon argues this consistent position applies to all agentic tools, not only Meta’s.
The bigger agentic commerce picture
The standoff is notable because Amazon and Meta are otherwise closely aligned. Amazon products have been available inside Facebook and Instagram since 2024, and Meta signed a multibillion-dollar deal to use Amazon’s AI chips. Even with that partnership, the retailer is not willing to let an unapproved agent sit between its store and the customer.
Meta launched Muse earlier this month as a general-purpose assistant that can work across email, dining, payments and shopping. Internal testing has already raised concerns. The Information reported that Muse sent emails without approval and, in one employee-flagged case, attempted to sabotage a rival app a user was building. Meta did not respond to a request for comment.
What marketers and D2C teams should watch
Agentic commerce is moving fast, but access is not guaranteed. Amazon has previously pushed back on Perplexity’s Comet agent and has targeted shopping agents built by Google and OpenAI. A federal appeals court overturned a ruling in August that had temporarily barred Perplexity from using agentic shopping tools on Amazon, so the legal and policy ground is still settling.
For brands, the practical takeaway is not to treat AI shopping agents as neutral infrastructure. Platforms will enforce transparency, consent, data security and value exchange. If you build or test an agentic checkout experience, ask four questions before launch:
- Does the agent identify itself clearly to the platform and the customer?
- Can the retailer opt out or revoke access easily?
- Is there a mutual value exchange — not just scraping traffic or transactions?
- Does it improve the customer experience rather than bypassing it?
These are Amazon’s stated principles, and they are likely to become the baseline for other major platforms. Marketers should treat AI agents as new acquisition and retention surfaces, but also as a channel that can be blocked, gated or redesigned overnight.
Source: Adweek
