In May 2026, Alipay announced that the number of AI payment transactions had exceeded 300 million. One month later, WeChat opened access to mini-program AI integration for developers, one requirement of which sparked controversy: developers must authorize the platform to read the source code of mini-programs.
The two time points are less than 30 days apart, but behind them are two diverging paths that have developed for over a year. According to LatePost, Alipay is internally testing an AI version code-named "Treasure Plan," which does not just add an assistant but allows users to switch to an entirely new, conversation-driven interface with one click. WeChat, on the other hand, has its president, Liu Chiping, setting the tone during a financial report call: it will eventually feature an AI agent, deeply connected to social relationships, official accounts, and video accounts, but there is no separate timetable.
The two platforms, each with 1 billion users and millions of mini-programs, have given opposite answers to the same question: when AI can perform services for users, should the entry point be rewritten or hidden?
Alipay is cutting more than just the interface
To understand what Alipay is doing, one must first look at a specific user action.
In the past, ordering three low-sugar lattes in Alipay and calling a car to the airport involved a standard process: find the Didi mini-program entrance, input the destination, and confirm the ride; exit and find the Luckin mini-program entrance, choose the type, adjust the sugar level, add to the cart, and settle the payment; switch back and forth between the two mini-programs to complete the payment. Each step involves a click, a page jump, and a wait.
The change that the "Treasure Plan" aims to implement is compressing this entire series of actions into one sentence. The user says to the chat interface, "Help me order a car to the airport and, by the way, reserve three low-sugar lattes nearby," and the AI takes over all subsequent steps: understanding the intent, breaking down tasks, calling corresponding transportation and dining services, combining orders, and completing payment. The interaction interface is no longer a row of mini-program entrances but a chat window.
This thorough change can be seen from the internal product design process. According to LatePost, to determine the new interaction form, the project team created over 100 product design versions. They ultimately chose a dialogue-centered solution based on the judgment that natural language has become the mainstream method for AI interaction, and service distribution should be rebuilt in this direction rather than sticking an AI patch on the old framework.
This radical approach was not Alipay's choice from the start. In the second half of 2023, when the Alipay business group management initiated discussions on "how to transition to intelligence," the first question that arose was: should it transform the original end or start a new one? Initially, the project team chose the latter. At the Bund Conference in September 2024, Alipay launched the independent AI application "Zhi Xiaobao," positioned as an AI lifestyle manager.
However, Zhi Xiaobao did not take off. According to insiders, the daily active users of the independent app are far less than the internal intelligent assistant. The dialogue assistant retained within Alipay, leveraging traffic from the homepage, has stabilized at a level of several million daily active users, accumulating far more substantive interaction data than the independent app.
There was also a more realistic constraint: at that time, Ant Group was focusing on the health app "Antifortune," and the general AI "Lingguang" was also being promoted, with limited computing power and development resources. Creating another independent app not only meant competing for resources with these projects but also carrying the enormous cost of migrating users from scratch.
In March 2025, the team pivoted and no longer insisted on the independent path. An internal judgment gradually formed: providing quality service to Alipay's existing 1 billion user base, allowing users to access AI services with zero migration cost, is more effective than trying to create a new application from the outside. In December 2025, the AI version of Alipay project was officially established, with the earliest members coming from the internal intelligent assistant team, later joined by algorithm, C-end product, and mini-program business teams.
The final product route is not an independent native app or embedded assistant within the existing application, but a one-click switch. After the new version went live, the default opened is still the original Alipay, and users can set the AI version as their preferred interface. LatePost reported that this "leave room for flexibility" approach points to an internal saying called "腾笼换鸟" (replacing old with new).
WeChat does not let AI obstruct interactions between people
WeChat's AI path has taken a different logic from the very beginning.
Tencent President Liu Chiping's statements during the Q3 2025 financial report call leave little ambiguity: the AI agent that WeChat plans to launch will deeply connect with social relationships, communication capabilities, official accounts, and video accounts, making it a unique agent. There is no aggressive timetable, and the official party has twice publicly debunked rumors about an AI assistant.
Why can't WeChat cut out a dialogue interface like Alipay? The reason lies not in technical capability but in product attributes. WeChat's core interface is the chat list, which is the most frequently opened mobile page for 1 billion people daily. Any attempt to add an AI conversation entry point on this interface might be seen by users as an interference in social relationships. Alipay's homepage is a service entrance; transforming this service entrance into a conversation window requires users to adapt to a new operating habit. WeChat's homepage is a dialogue between people; replacing or displacing that with AI dialogue touches on the most crucial aspect of user psychology.
WeChat's plan is closer to a "parasitic" logic. The AI assistant does not replace any interface; it hides in group chats, official accounts, waiting to be summoned as an agent. Imagine a scenario in a family WeChat group where someone forwards a lengthy article about family camping spots, and other members do not need to click to read; they can directly ask the AI assistant to summarize the key points and coordinate the group members' calendars to book the recommended campground mentioned in the article. The agent digests the official account's content, calls reservation services in mini-programs, coordinates timing based on the schedule information of multiple group members, and finally pushes the reservation results back to the group.
Throughout the process, AI operates within the context of the group chat, and what users see are still this group, these people, and these conversations. The "services" the agent completes are embedded in social relationships rather than popping up in a new interface to showcase its presence.
This restraint comes with its cost. In WeChat, services are presented in the form of mini-programs on the platform, numbering in the millions. To let AI handle these tasks for users, it needs to understand not only user intent but also the data structure, page logic, and interaction processes of these services themselves. Alipay faces the same issue, and the solutions from the two represent the most crucial divergence in this sector.
Reading screens and reading source code, which solution is harder
In June 2026, the WeChat open community released the "Mini Program AI Development Mode (beta) Access Guide," which provides two modes.
The first is the "Automatic Mode." Developers authorize the platform to read the mini-program source code during the review process. The AI understands the page structure and operation logic by analyzing the source code and directly manipulates the mini-program. The second is the "Development Mode," where developers package their services into Skills according to WeChat's defined protocol, including atomic interfaces and atomic components, enabling the AI to complete tasks by calling these standardized interfaces.
Alipay's solution is a "dual-track system." According to LatePost, on one side, it encourages willing merchants to actively integrate, turning their services into AI-accessible MCP or Skills; on the other side, under user authorization, the AI operates through "screen reading" of existing mini-program interfaces to accommodate services that have not yet been transformed.
When compared, the core difference lies in: when transforming unprepared existing mini-programs, WeChat requires developers to hand over their source code, while Alipay chooses to let AI perform visual operation for users.
From the WeChat open community documentation, it appears that "Automatic Mode" is technically the more thorough solution. After the AI reads the source code, its understanding of the page is structured, the operational path is clear and controllable, unlike screen reading that relies on visual recognition and interface simulation, thus having a lower probability of errors. However, this solution shifts the pressure to developers. The source code is the core asset of mini-program developers; handing it over means fully exposing their business logic, data structure, and interaction designs to Tencent. For small to medium-sized businesses that rely on mini-programs for their business, this is not just a security concern but a commercial risk: once the service processes are entirely controlled by the platform, how much space remains for traffic distribution and bargaining?
If developers do not choose the "Automatic Mode," the development mode is not easy either. Developers need to rework their business processes into atomic capabilities, encapsulate them into Skills according to WeChat's defined protocol, and go through a new review process. For a restaurant mini-program encompassing ordering, payment, coupon redemption, and membership points, the effort required to deconstruct and package could be several times that of initial development. Who will bear this cost? WeChat has not provided any incentive plans, at least for now.
Alipay's screen reading solution bypasses these issues. It does not require merchants’ cooperation, does not need code changes, and merchants do not even need to know that their mini-program is being operated by AI. The user simply tells the dialogue interface, "Help me buy a train ticket to Shanghai," and the AI opens the 12306 mini-program interface, identifies the departure point, destination, train list, seat selection buttons, and payment confirmation page, simulating the user's finger operations step by step. For merchants who have already completed MCP or Skill integration, AI can directly call standardized interfaces for a smoother experience; for a vast number of untransformed long-tail services, screen reading provides a minimum threshold compatible path.
However, the issue with screen reading is also straightforward: its stability has not been verified on a large scale. The interfaces of mini-programs vary greatly, and dynamic loadings, pop-up ads, and layout variations due to version updates can all increase the likelihood of AI recognition failures. If the position of a payment confirmation button shifts by a few pixels, can AI accurately hit the mark? If there is an operational error during the screen reading process, such as mistaking the amount or selecting the wrong delivery address, who bears the responsibility? Alipay has not disclosed relevant exemption clauses and dispute handling mechanisms.
The logic behind this path is to let users use it first. Once merchants see the order conversions brought by AI, they will naturally integrate the standard interface to optimize the experience. B-end is forced by C-end.
What 300 million transactions have verified
Apart from products and ecosystems, Alipay has done something else related to how to pay with AI.
At the AI payment ecosystem conference in May 2026, Alipay revealed that the number of AI payment transactions had exceeded 300 million, supporting 95% of general intelligent agent frameworks, and also launched Token Pay and AI Wallet. These two products are key to understanding the infrastructure of the Agent economy.
Token Pay solves the problem of extremely small amounts and high-frequency payments. When AI compares prices between two takeout platforms, it might need to call a 0.01 yuan verification transaction to confirm the account's validity; when AI filters the best combination of multiple coupons, verifying each coupon constitutes a payment action. These transaction amounts are small, but their frequency far exceeds that of human users. The previous payment system was designed for "human confirmation, human payment," while Token Pay assigns this action to the agent.
AI Wallet is more like giving an agent a budget card. Users set rules and limits, and the AI autonomously completes payments within those rules. Ant Group CEO Han Xinyi posited at the conference that in the future, countless agents might be active in economic activities, with interactive actions transitioning from human-to-human interactions to human-to-agent interactions, as well as interactions between agents.
While the absolute value of 300 million does not seem large relative to Alipay's overall annual transaction scale, its significance lies in verifying one thing: users have allowed AI to complete real commercial transactions on their behalf, not just stopping at inquiry and price comparisons. From a sentence to call a car and order food to AI making payments, the technological and user authorization chains of this service loop have been connected.
WeChat Pay has not yet publicly disclosed a specific plan for AI transformation. WeChat Pay also covers a massive user base, but its scenarios are more associated with social transfers, red envelopes, and merchant payments. The form of the agent economy may differ; whether both will form new discrepancies in payment infrastructure depends on whether WeChat’s AI assistant is launched alongside similar Agent payment capabilities.
The ecosystem is being torn open along two cracks
Both Alipay and WeChat point towards agent service entry points, but the differences in their paths will tear open two different rifts in the mini-program ecosystem.
Alipay's screen reading solution passively automates many long-tail mini-programs with AI. Merchants have done nothing, yet users can already operate their services through AI. This will produce two reactions: some merchants find that the order volume brought by AI is rising and actively integrate MCP or Skills to optimize the experience and seek more traffic distribution; others may become resistant, as the source of orders becomes unclear. Previously, each click by users within the mini-program was traceable; now, the path during AI's screen operation means merchants cannot access user behavior data.
Internally, Alipay has evidently anticipated this. LatePost reports that after the AI version of Alipay goes live, an AI open platform for merchants and developers will be launched quickly. This platform will likely address how to allow merchants to enjoy increased orders from AI while retaining visibility and control over service processes, user outreach, and revenue distribution.
The pressure on WeChat's side is different. The source code authorization threshold will divide developers into two groups. Top developers, with technical teams and negotiating leverage, may be willing to hand over source code or invest resources in packaging Skills for priority traffic distribution from WeChat AI assistants. However, many small and medium-sized merchants might neither be willing to hand over their source code nor able to bear the packaging costs. If WeChat AI assistant launches and traffic indeed leans towards authorized merchants, those unauthorized mini-programs may become marginalized in the distribution channel for AI services. Over time, WeChat's mini-program ecosystem could further concentrate towards the top, contradicting WeChat's long-stressed narrative of a "decentralized" ecosystem.
A more subtle issue lurks in the technical standards. Alipay promotes MCP and WeChat has defined its own set of mini-program MCP protocols. Although the names are the same, specific implementations are not fully interoperable. A restaurant merchant wishing to enable both Alipay AI and WeChat AI to call their ordering service may need to package it according to two sets of specifications. This is not a technical challenge, but it is a cost issue. Which side secures scale advantages first will have greater bargaining power to push industry standards. At this point, with AI payments from Alipay exceeding 300 million transactions, this advantage temporarily rests with Alipay.
Returning to the user end, the ultimate result of the transformation may redefine the relationship between people and their phones. If Alipay's dialogue interface is successful, the frequency and scenarios in which users open Alipay will change. It will not just be opened when making payments, but rather whenever there is a need to casually ask something. If WeChat's agent succeeds, the way users accomplish tasks in group chats will change. There will be no need to exit the chat interface to find services; everything will be completed through the agent within the group chat.
The "red envelope war" between the two platforms on the eve of the 2014 Spring Festival altered which account users preferred to hold their money in. This time, the contention is over who gets to be entrusted with the phrase "help me get this done."
12 years ago, WeChat's red envelope was called a "Pearl Harbor attack" by Jack Ma. 12 years later, as fake and real news about WeChat AI circulated for months, Alipay took the lead on the stage. Which of the two paths is more aligned with the real needs of the agent era will not be found in product launch events but in how millions of mini-programs are reactivated and in the experiences of billions of users saying "help me" to their phones for the first time.
免责声明：本文章仅代表作者个人观点，不代表本平台的立场和观点。本文章仅供信息分享，不构成对任何人的任何投资建议。用户与作者之间的任何争议，与本平台无关。如网页中刊载的文章或图片涉及侵权，请提供相关的权利证明和身份证明发送邮件到support@aicoin.com，本平台相关工作人员将会进行核查。
