Alipay AI Pay processed 120 million transactions in a single week in February 2026. By May, the total had crossed 300 million. Over 100 million users were already engaged. These are not projections or pilot metrics—they are operational numbers from a working system.
The numbers matter because they challenge a core assumption in Western agentic commerce discussions: that scaled, trust-verified agent transactions remain a future-state problem. The industry has been debating which protocol will unlock ‘Category 3’ commerce—high-trust consumer transactions where an AI agent can search, compare, authorize, pay, and manage follow-up actions with minimal human intervention. Meanwhile, Alipay has built the infrastructure to run it.
The core of this development is the Agentic Commerce Trust (ACT) Protocol, launched in January 2026. Unlike the fragmented approaches seen elsewhere, ACT serves as an open standard connecting diverse partners including the Qianwen App, Taobao Flash Deals, and Alibaba Cloud Bailian. This is complemented by the AI Wallet, which provides the necessary consumer-facing controls—permissions, monitoring, and transaction reviews—that address the inherent risks of autonomous spending. As Peng Yang, CEO of Ant International, noted, ‘New models of financial interoperability and AI commerce give us powerful tools to help global giants as well as mom-and-pop shops to increase resilience and expand revenue streams.’
The Western landscape, by contrast, remains defined by protocol fragmentation. Google’s UCP, Mastercard’s AP4M, and Visa’s Intelligent Commerce—which includes an Agent Score and Agentic Directory—are significant efforts, yet they operate within a market that relies on coordination across multiple, often disconnected, applications. Even x402, while reporting over 100 million transactions on Base, faces scrutiny regarding volume quality. Artemis Analytics estimated in February 2026 that approximately 50% of x402 activity was wash trading. Stripe’s Shared Payment Tokens (SPTs) represent a buyer-side approach to scoped permissions, but they remain one piece of a larger, unintegrated puzzle.
This structural divergence shows up in consumer behavior. Euromonitor’s March 2026 data indicates 23% of Chinese consumers now purchase directly through AI, compared to 15% in the US and 12% in the UK. The gap is not purely technological. A Product.ai survey from April 2026 found that only 14% of US consumers trust AI recommendations without verification—86% insist on reviewing purchases first. Forrester’s June 2026 report highlights that trust and adoption remain uneven across demographics, suggesting the ‘Category 3’ absence is less about missing technology and more about missing consumer confidence in the current Western framework.
The difference, as Euromonitor frames it, is that ‘China is moving towards a world where the AI and the commerce are the same thing, while the West is using AI-shaped commerce but not yet to replace existing structures.’ The US model remains fragmented, relying on coordination across multiple apps while giving retailers less control over the underlying LLM infrastructure.
Western efforts are not failing—they are pursuing a different strategic objective. Karan Katyal, Global Head of Agentic Commerce at Adyen, articulated this philosophy: ‘We believe the future of agentic commerce should be open, so we intentionally designed Adyen Agentic to help retailers integrate once and participate across evolving platforms, protocols, and experiences — without having to bet on which ecosystems ultimately win.’
The implications are straightforward. If the goal is to achieve the scale seen in the Alipay ecosystem, protocol-agnostic, multi-rail integration may face challenges providing the seamless, high-trust environment consumers currently demand. The debate over whether agentic commerce is a future-state or a present-day reality is no longer hypothetical. The infrastructure is being tested at scale, and the results suggest the path to adoption may be paved by integration rather than protocol competition.