The milk tea purchased from that store yesterday; please place another order for me.

"The milk tea purchased from that store yesterday; please place another order for me."

If one day you could truly say this to your phone, and it would find the merchant, confirm the product, place the order, make the payment, and arrange delivery on your behalf, you would probably not first think of a payment revolution. Most likely, you would simply feel that your phone has finally begun to act like an assistant capable of getting things done.

This may well be the more appropriate starting point for today’s discussion of AI-powered payments.

For most people, AI-powered payments do not mean giving AI a wallet or letting robots decide how to spend money on their own. What users can truly perceive is whether a single sentence can be directly converted into a transaction. You state your need, and the system completes the subsequent steps for you.

Therefore, what is truly worth discussing about AI-powered payments is often not the word “payment” itself, as that term is overly abstract and distant. For the broad base of internet users, the prevailing experience may be this: when I express my shopping needs in plain language to an app or an AI assistant, it can handle the entire process directly.

 

Chinese Users’ Intuition

If we look solely at the Chinese market, AI-powered payments are in fact not so unfamiliar.

The reason is simple. Chinese users have long lived within a highly mature mobile internet ecosystem, where food delivery, e-commerce, mapping, ride-hailing, local lifestyle services, mobile payments, and instant delivery are inherently interconnected. In many cases, what users want is not an entirely new payment method, but rather to avoid switching back and forth between apps and repeating the same clicks.

In China, AI-powered payments resemble more of an upgrade to the e-commerce shopping experience than a ground-up reconstruction of the payment system.

When Alibaba upgraded the Qwen App in January of this year, the direction it set was quite typical. According to the official introduction, Qwen has integrated services within the Alibaba ecosystem, including Taobao, Taobao Flash Sales, Alipay, Fliggy, and Amap. With a single sentence, users can order meals, complete payments within chat interfaces, and plan and book itineraries. It may appear that AI has become smarter, but more accurately, the platform’s previously integrated capabilities have now been reorganized.

This is also why, in China, the most viable applications of AI-powered payments are not grandiose visions, but rather highly routine, everyday scenarios.

For example, "Buy another box of the printer paper I purchased last week and deliver it to the office." Or, "Place another order for the takeout from that restaurant yesterday and have it delivered to my home." Another example is, "Book a restaurant close to the company, suitable for four people, with a budget under 500 yuan."

Users understand these requests immediately. This is because the AI here does not resemble some mysterious new financial instrument; it is simply a smarter shopping assistant.

Chinese internet giants will pursue whatever the public demands, amid intense competition. Currently, there are roughly three implementation paths for AI-powered payments.

One approach is to let the AI operate on-screen on your behalf, clicking buttons, filling in information, and submitting processes just like a human. The advantage of this method is that it can be tested anywhere; the disadvantage is also obvious: it is prone to errors when page layouts change, and key steps often still require your final confirmation. An example is the Doubao phone, which faced restrictions from WeChat shortly after its launch.

Another approach is for platforms to directly expose their service capabilities for invocation by AI. For Chinese tech giants, this path is actually the most straightforward. Because payment processing, merchant services, mapping, order management, and delivery networks are already within their own ecosystems, the AI does not need to mimic user clicks; it can directly call upon system capabilities. From the user's perspective, this means, "I say one sentence, and it actually gets the job done." Examples include Alibaba's Qwen and ByteDance's Doubao.

The third approach is more difficult, involving cross-platform, cross-website, and cross-company operations. At this stage, the issue is no longer just about product usability, but whether other parties are willing to allow your AI to act on behalf of the user. Currently, there is no indication that any Chinese internet company is attempting to challenge this "hard mode."

Today, China's clear advantages lie in the first two approaches, especially the second one.

This is why, for Chinese users, AI-powered payments may not seem particularly groundbreaking. The overall perception is likely not that "AI payments have arrived," but rather that "mobile shopping has become more convenient."

 

The situation overseas is different

However, once the perspective shifts to overseas markets, things become somewhat different.

Overseas markets certainly have e-commerce, food delivery, mapping, and payment services, but they lack the highly integrated platform structure found in China. Service entry points, merchant systems, payment tools, and fulfillment networks are often controlled by different companies. Therefore, when AI attempts to complete a transaction on behalf of a user, the challenge is not merely "whether an order can be placed," but "whether others recognize your AI agent."

This is why AI-powered payments overseas have suddenly become more complex.

A typical example is the conflict between Amazon and Perplexity. Perplexity’s Comet browser positions AI-powered shopping as a key selling point, with its official description stating that it can help users compare products, read reviews, and proceed all the way to checkout. This aligns closely with many people’s most intuitive conception of AI shopping: instead of browsing website by website, users let the AI select and purchase items on their behalf.

Yet this is precisely where the problem arises.

According to court filings dated March 9, 2026, a core argument advanced by Amazon in the litigation is that Perplexity’s AI agent accessed user accounts and attempted to perform actions in circumstances where “the user consented, but Amazon did not.”

Amazon’s refusal stems from reasons largely consistent with those underlying WeChat’s disagreement with Doubao Mobile: your authorization to an AI does not equate to the website’s authorization to that AI. Your willingness to let it make purchases on your behalf does not mean the platform is willing to allow a third-party agent to enter its systems and initiate transactions on your behalf.

While the public rationale is framed around considerations of customer privacy and security, the core concern is that AI-powered shopping would directly divert billions of dollars in advertising expenditures paid by sellers on its platform.

After all, AI shopping tends to be highly utilitarian: it goes straight for the desired product and does not view advertisements.

In China, such conflicts are often absorbed by platforms’ integrated capabilities; however, overseas, once AI needs to complete transactions across websites, merchants, and payment networks, these issues become unavoidable.

Thus, in overseas markets, AI payments are not primarily a question of “usability,” but rather one of “whether an end-to-end closed-loop process can actually be completed.”

 

Approaches to AI Payments in Overseas Markets

This is why Google, Stripe, and Coinbase have recently been focusing their efforts on this issue.

They recognize the same underlying challenge: in traditional internet transactions, the default assumption is that the person placing the order is a human sitting in front of the screen, personally browsing, confirming, and paying. Once the ordering party becomes an authorized AI, that default premise no longer holds. How can a website verify that the AI was indeed dispatched by you? How can a merchant confirm that it is not making unauthorized purchases? And how should payment institutions determine whether the funds ought to be released?

Google The initial focus is on resolving the foremost trust issue. In September 2025, Google launched AP2 not to teach AI how to make payments, but to answer a fundamental question: when an AI initiates a payment on behalf of a user, what basis does the system have to believe the transaction is authentic? Google’s approach is to attach a set of verifiable “proofs of authorization” to such transactions. Within AP2, it designed two critical components: one called Cart Mandate, which can be understood as the user’s signed confirmation for a specific purchase; and another called Payment Mandate, intended for payment networks and card issuers, indicating that the transaction is initiated by an agent, whether the user was present at the time, and the context of the transaction. Google later integrated this framework with PayPal to create a more comprehensive merchant solution: merchants can use their own conversational shopping assistants to engage users, and at the payment stage, PayPal Agent takes over to complete the authorization and payment processes using the AP2 mechanism. In essence, Google’s path is focused first on establishing “why AI-initiated orders should be trusted.”

Stripe approach aligns more closely with merchants and platforms. It essentially operates on two levels. The first level is the Agentic Commerce Protocol (ACP), launched jointly with OpenAI in September 2025. You can understand this as a set of “AI-readable checkout standards.” Rather than having AI mimic human behavior by clicking through web pages and filling out forms, Stripe encourages merchants to proactively expose capabilities such as product listings, inventory, and checkout functions, allowing AI to initiate purchases through standardized interfaces. The advantage of this approach is that merchants remain primarily responsible for orders and fulfillment, retaining control over product display, order processing, and risk management, without needing to rebuild their systems separately for each AI platform. Stripe subsequently launched the Agentic Commerce Suite, aiming to provide one-stop tools for catalog integration, checkout, and payment, thereby lowering the barrier to entry for merchants.

However, Stripe has identified another entirely different scenario. This is not about “AI helping people buy things,” but rather “software paying software directly.” Consequently, in March 2026, it launched Machine Payments. In this scenario, AI is not purchasing a cup of cola, but rather buying APIs, data, computing power, content access rights, or even paying per invocation for specific services. Stripe’s solution allows merchants to convert their interfaces into pay-per-use models, accepting amounts as low as 0.01 USDC, with funds deposited into Stripe accounts and ultimately settled and reconciled through Stripe’s familiar processes. For agents, this eliminates the need to register accounts, apply for API keys, and navigate numerous manual procedures; instead, they can pay as they invoke services. In other words, while Stripe is building checkout infrastructure for AI shopping, it is also establishing a track for small-value payments between machines.

Coinbase perspective differs again. It proceeds from the assumption that many future AIs will not be “shopping assistants” but rather “software with wallets.” Therefore, in February 2026, it launched Agentic Wallets. The core objective is not to make AI better at browsing websites, but to enable AI to truly possess a wallet capable of spending, receiving, and transacting funds, complete with security boundaries. Within Coinbase’s framework, components include the wallet infrastructure itself, machine payment protocols such as x402, and pre-packaged capabilities such as top-ups, transfers, transactions, and yield management. Its primary focus is addressing a different set of issues: if an AI wishes to purchase computing power, buy data, pay API fees, or execute certain automated transactions on-chain, can it do so without requiring human confirmation each time? Base’s official documentation refers to such agents as “independent economic actors.” While this terminology may sound weighty, the meaning is clear: Coinbase aims not merely to make AI resemble a shopping assistant, but to enable AI itself to become a software entity capable of managing income and expenditures, settling accounts, and conducting transactions.

When these three approaches are viewed together, the differences become clear.

Google seeks first to address “how you prove that a transaction was truly authorized by me.” Stripe is more concerned with “how merchants can expose their products and checkout capabilities to AI.” Coinbase takes a step further, addressing “if the payer is software itself, how it can independently manage income and expenditures.”

Precisely for this reason, although all three appear to be developing AI payment solutions, their focal points differ. Google is effectively adding a layer of trust mechanisms for cross-platform agency transactions; Stripe aims to integrate both AI shopping and AI service invocations into existing commercial networks as much as possible; Coinbase is betting more heavily on a world where software pays software directly.

 

Two Narratives

At this point, the distinctions become clear.

In China, AI payments resemble a natural extension of the mature internet ecosystem. Platforms have already organized merchants, payments, logistics, mapping, and local life services; AI’s role is to reorchestrate these capabilities, shifting users from “manual operation” to “simple voice commands.”

Overseas, AI payments resemble a belated effort to catch up. Because no single platform naturally controls the entire chain, once AI attempts to truly complete transactions on behalf of users, it immediately encounters practical issues such as website access permissions, merchant acceptance, recognition by payment institutions, and allocation of liability in case of problems.

In China, the focus is more on advancing the internet shopping experience by one step.

Overseas, the focus is more on renegotiating the rules of participation among various parties in the context of internet software and online consumption scenarios.

Precisely for this reason, Chinese users’ initial reaction to AI-powered payments may be, “This feature is quite convenient,” whereas overseas markets are more likely to perceive AI-powered payments as signaling that “the internet’s original transaction methods may need to change.”

Let us return to the opening scenario.

“Place another order for the milk tea I bought at that shop yesterday and have it delivered to me.”

What is truly noteworthy about this statement lies not in the word “payment,” but in the fact that it delegates many tasks that previously required manual execution to the system. The system must identify which store, which product, what specifications, the delivery address, and the payment method. While payment is certainly important, it is merely the most inconspicuous step in the entire process.

Therefore, the focus of AI-powered payments is not payment itself.

What people truly care about is not which protocol is used in the backend, whether settlement occurs on-chain, whether stablecoins are employed, or whether cryptographic algorithms are utilized. Ultimately, people will judge its utility by a simple criterion:

When I give an instruction to the AI, can it actually complete the task on my behalf?

 

Recommended Courses

The advent of the AI-powered payment era is not an issue confined to any single link—from regulatory characterization and licensing pathways, to the establishment of anti-money laundering (AML) frameworks and off-chain and on-chain risk control, and further to compliance implementation in real-world business scenarios. Each link directly affects whether AI agents can truly “emerge.”

Everything is new. Drawing on Mankun Law Firm’s long-standing expertise in Web3 legal practice, we are pleased to launch the “Crypto Payment Compliance: Practical Closed-Door Workshop (Shenzhen Session).”

April 18–19, in Shenzhen. Limited seats in a small-class format, offering multi-perspective insights from legal, compliance, and security angles.

Instructors include representatives from Mankun Law Firm, Chainalysis, Beosin, KUN Kun, COBO, as well as practical leaders from duly licensed institutions.

Coverage includes: regulatory pathways and licensing frameworks, compliance governance structures, anti-money laundering (AML) practices, on-chain security, and frontline insights from licensed institutions.

If you are developing crypto payment-related products or considering the compliance pathway for AI Agents in payment scenarios, we recommend scanning the QR code to consult with us.

 

Author

Liu Honglin, Founder of Mankun Law Firm. Member of the Young Lawyers Working Committee of the Shanghai Bar Association, Member of the Information Technology Working Committee of the Shanghai Bar Association, and Member of the Legal Technology Working Committee of the Shanghai Bar Association. Lawyer Liu Honglin has 10 years of experience in law and internet entrepreneurship. He previously served as Vice President of a legal technology company under Tencent’s strategic investment arm and as Legal Manager for a private equity fund at a listed company. He specializes in proposing practical, actionable solutions for cases from the perspectives of business models and legal practice, thereby maximizing commercial benefits for clients.

 

About Mankun

Mankun Law Firm was established in 2015 as a boutique law firm in China focusing on the Web3 new economy and deeply engaged in the blockchain industry. The Mankun team possesses unique and diverse industry backgrounds, with members hailing from renowned legal service providers, state judicial organs, internet technology companies, crypto asset institutions, and blockchain industry think tanks.

Leveraging a profound understanding of the new economy, continuous attention to and research on policies and regulations, and extensive practical experience, the Mankun team is adept at providing comprehensive legal services from the perspectives of business models and legal practice. These services include business structure design, project financing and investment, transaction planning, operational compliance, resolution of complex civil and commercial disputes, prevention and control of criminal risks, and criminal defense for new economy enterprises in sectors such as Web3, blockchain, AI, NFTs, digital collectibles, crypto funds, crypto payments, DeFi, real-world assets (RWA), and GameFi.

Mankun Law Firm is headquartered in Shanghai, with branch offices in Hong Kong (China), Silicon Valley (USA), Shenzhen, Hangzhou, Zhengzhou, and Chengdu. To meet the global compliance development needs of Web3 industry clients, Mankun has established local offices in major global crypto-financial hubs and selected local professional blockchain service partners, providing clients with professional legal and compliance services that combine global breadth with deep expertise in China.