China Already Fought the AI Shopping-Agent War. Europe Is Next.
Set Trending Topics as a preferred source on Google.
Amazon has blocked Meta’s new AI agent Muse, and China saw the same conflict nine months earlier. In this guest post, Ting Wasner-Lian explains how it is being resolved there and what that means for Europe’s platforms, retailers and advertisers.
On 8 September 2026, Meta launched Muse, a personal AI agent that can shop and pay on the user’s behalf. On 21 September, Amazon blocked Muse, arguing that the agent had accessed the retail platform without authorization or proper disclosure, violating its terms of service and bypassing Amazon’s control over customer data and monetization.
Anyone who follows the Chinese market has seen this before. In December 2025, ByteDance released its first AI phone with its Doubao AI agent built in. Within days, Taobao, WeChat and several banking apps blocked it, and ByteDance had to switch off part of what the agent could do.
Both cases look like technical disputes about AI agents and security. But I think something else is at stake: a fight over who owns the moment of purchase.
Why China and the US Build Personal AI Agents So Differently
The Doubao phone operates the user’s own phone and taps through apps the way a person would. Muse works from its own dedicated virtual machine in Meta’s cloud, which acts as a personal computer that keeps working even after you close the app.
The reason reflects two different internets. US services sit largely on the open web and on APIs, while Chinese commerce and services are concentrated in closed super apps such as WeChat, Taobao and Meituan. Many tasks cannot bypass the native app, so the agent has to live on the phone.
Yet They Face the Same Problems
Platforms such as Taobao and Amazon are paid for attention: search results, sponsored products and recommendations. When an agent does the shopping for you, nobody sees those pages anymore, and the advertising revenue disappears. The security concerns platforms cite are real, but blocking also protects their business and buys time for their own agents, such as Amazon’s Alexa for Shopping.
I believe search ads and sponsored listings will be hit hardest. Social feed ads, which create new demand, are safer. That is why Meta may end up the winner.
From Blocking to Bargaining
China shows what comes next. After the blocking came negotiation: from January 2026 on, ByteDance integrated third-party ride-hailing, food delivery and ticketing services directly into its Doubao AI ecosystem. In September, it introduced a formal access protocol called SAEP. After 15 October, apps that have not explicitly opted out will be opened to the agent step by step.
The US is moving down the same path, only faster. One day after Amazon’s block, Meta and Shopify announced a deal that lets Muse buy across Shopify’s merchant network.
My take: no agent will be able to shop everywhere any time soon. Platforms will run their own agents behind their own walls, and outsiders will get in only through negotiation.
Business Models Shift From Selling Attention to Charging for Outcomes
Until now, advertising has bought attention in order to lead to a sale. When agents complete the sale directly, platforms will charge for the result instead: commissions per transaction, access fees for outside agents, subscriptions and more. Paid placement inside agent recommendations is possible but risky, because an agent that is secretly paid loses its users’ trust.
What Europe Should Do Now
I expect Europe to take a third path. In China, super apps decide what AI agents may do. In the US, commercial deals do. In Europe, regulators will, through rules on labelled rankings, data protection and gatekeeper platforms.
For retailers, the most urgent task is making product data readable for agents, with accurate prices, stock levels, delivery and return terms, well before thinking about advertising to them. China has shown that this fight is about who controls the purchase. Europe gets to watch it play out twice before it arrives here, and it should use that time.
About the Author
Ting Wasner-Lian is the founder of SE Incubator Consulting in Vienna and helps European decision-makers understand Chinese AI.
