In May 2026, Alipay announced that the number of AI payment transactions had exceeded 300 million. A month later, WeChat opened up AI access to its mini-programs to developers, one of the requirements of which sparked controversy: developers must authorize the platform to read the mini-program's source code.

The two time points are less than 30 days apart, but they represent two diverging paths that have been diverging for over a year. According to LatePost, Alipay is internally testing an AI version codenamed "Bao Plan," which isn't about adding an assistant, but rather allowing users to switch to a completely new, conversation-driven interface with a single click. WeChat, on the other hand, through its president Martin Lau in its earnings call, stated that it will eventually feature an AI agent, but with deep integration with social relationships, official accounts, and video accounts, without a separate timeline.

Two platforms with 1 billion users and millions of mini-programs have given opposite answers to the same question: When AI can operate services for users, should the entry point be rewritten or hidden?

Alipay cut out more than just the interface.

To understand what Alipay actually does, we need to look at a specific user action.

In the past, the standard process for ordering three low-sugar lattes and calling a car to the airport through Alipay was: find the Didi mini-program, enter the destination, and confirm the ride; exit, find the Luckin Coffee mini-program, select the product category, change the sugar content, add to cart, and checkout; switching back and forth between the two mini-programs to complete the payment. Each step involved a click, a page jump, and a wait.

The change that "Bao Plan" aims to make is to compress this entire process into a single sentence. The user says to the chat box, "Call me a car to the airport, and order three low-sugar lattes nearby," and AI takes over all subsequent steps: understanding the intent, breaking down the task, calling the corresponding travel and dining services, combining the order, and completing the payment. The interface is no longer a series of mini-program entries, but a chat window.

The thoroughness of this change is evident from the internal product design process. According to LatePost, the project team went through over 100 product design versions to determine the new interaction model. The final choice of a dialogue-centric solution reflects the judgment that natural language has become the mainstream method of AI interaction, and service distribution should rebuild its entry point in this direction, rather than simply patching an AI element into the old framework.

This radical approach wasn't Alipay's initial choice. In the second half of 2023, when Alipay's management initiated a discussion on "how to move towards intelligentization," the first question they faced was: should they modify the existing platform or start a completely new one? The project team initially chose the latter. At the Bund Summit in September 2024, Alipay released its independent AI application, "Zhi Xiaobao," positioned as an AI life assistant.

Zhi Xiaobao failed to take off. According to sources, the standalone app's daily active users (DAU) are far lower than the in-app intelligent assistant. The conversational assistant, which resides within Alipay and leverages homepage traffic, has maintained a stable DAU of several million, accumulating far more interaction data than the standalone app.

There was another, more practical constraint: at the time, Ant Group was focusing its efforts on the health app "Ant Afu" and was also advancing the general AI "Lingguang," leaving it with limited computing power and development resources. Creating another standalone app would not only mean competing with these projects for resources, but also incurring the enormous cost of getting users to migrate from scratch.

In March 2025, the team changed course, abandoning the independent platform approach. An internal consensus gradually formed: serving Alipay's existing user base of 1 billion and providing users with AI services at zero migration cost was more effective than creating a new application outside the platform. In December 2025, the AI-powered Alipay project was officially launched, with the initial team coming from the in-app intelligent assistant team, later joined by teams from the algorithm, C-end product, and mini-program business teams.

The final product roadmap is neither a standalone native app nor an assistant embedded in an existing application, but rather a one-click switch. After the new version is launched, the original Alipay app will still open by default, but users can set the AI version as their preferred interface. LatePost reports that this "leaving room for maneuver" approach points to an internal term called "restructuring and upgrading."

WeChat prevents AI from standing between people.

WeChat's AI strategy follows a completely different logic from the very beginning.

Tencent President Martin Lau's statement during the Q3 2025 earnings call was almost unambiguous: WeChat's upcoming AI assistant will be deeply integrated with social relationships, communication capabilities, official accounts, and video accounts—a unique agent. There was no aggressive timeline, and the company has twice denied rumors about the AI assistant.

Why can't WeChat create a separate chat interface like Alipay? The reason isn't technical capability, but rather the product's nature. WeChat's core interface is the chat list, the most frequently accessed mobile screen for a billion people every day. Any attempt to overlay an AI chat entry point onto this interface is likely to be perceived by users as an interference with social relationships. Alipay's homepage is a service portal; turning it into a chat window would require users to readjust to a different operating habit. WeChat's homepage is about conversations between people; replacing or squeezing out human conversations with AI chat would infringe upon a crucial psychological territory for users.

WeChat's approach is closer to a "parasitic" logic. The AI assistant doesn't replace any interface; it's hidden within group chats and official accounts, acting as an agent waiting to be activated. Imagine this scenario: In a family WeChat group, someone forwards a long article from an official account about family camping sites. Other members don't need to open and read it; they can directly ask the AI assistant in the group to summarize the key points and coordinate with group members' calendars to book the campsite recommended in the article. The agent processes the content from the official account, calls upon the booking service in the mini-program, coordinates the time based on the schedules of multiple members in the group chat, and finally pushes the booking results back to the group.

Throughout the process, the AI operates within the context of the group chat, and the user still sees the group, the people, and the conversations. The "tasks" performed by the agent are embedded in the social relationships, rather than popping up on a separate interface to demonstrate its presence.

This restraint comes at a price. Within WeChat, services are hosted on the platform in the form of mini-programs, numbering in the millions. For AI to handle these tasks for users, it needs to understand not only the user's intent but also the data structure, page logic, and interaction flow of the services themselves. Alipay faces the same problem, and the two companies' solutions represent one of the most fundamental differences in this field.

Which solution is more difficult: screen reading or source code reading?

In June 2026, the WeChat Open Community released the "Mini Program AI Development Mode (beta) Access Guide", which provides two modes.

The first mode is "automatic mode." Developers authorize the platform to read the mini-program's source code during the review process. AI analyzes the source code to understand the page structure and operational logic, directly controlling the mini-program. The second mode is "development mode." Developers encapsulate their services into Skills according to WeChat's defined protocols, including atomic interfaces and components. AI completes tasks by calling these standardized interfaces.

Alipay's solution is a "dual-track system." According to LatePost, on one hand, it encourages willing merchants to actively integrate their services and make them into MCPs or Skills that AI can directly call; on the other hand, with user authorization, AI can "read" the screen of existing mini-program interfaces to make services that have not yet been modified compatible.

The core difference between the two lies in the fact that when modifying existing mini-programs that are not yet ready, WeChat requires developers to hand over the source code, while Alipay chooses to let AI interpret images and operate on behalf of users.

According to the WeChat Open Community documentation, the "automatic mode" is technically a more thorough solution. After AI reads the source code, its understanding of the page is structured, and the operation path is clear and controllable, unlike screen readers which rely on visual recognition and interface simulation, resulting in a lower probability of errors. However, this solution shifts the burden to the developers. Source code is the core asset of mini-program developers; handing it over means completely exposing their business logic, data structure, and interaction design to Tencent. For small and medium-sized businesses that rely on mini-programs, this is not only a security concern but also a commercial risk: once the platform has complete control over the service process, how much room for maneuver is left in traffic distribution and negotiation?

Even without selecting "automatic mode," development is not easy. Developers need to re-examine business processes, break them down into atomic capabilities, encapsulate them into Skills according to WeChat's defined protocols, and then go through a new review process. The workload of breaking down and encapsulating the entire process of ordering, payment, coupon redemption, and membership points for a restaurant mini-program could be several times that of initial development. Who will bear this cost? WeChat has not provided an incentive plan, at least not yet.

Alipay's screen reader solution bypasses these problems. It requires no merchant cooperation, no code modification, and merchants don't even need to know that their mini-program is being operated by AI. A user says to the chat interface, "Buy me a train ticket to Shanghai," and the AI opens the 12306 mini-program interface, recognizing the departure point, destination, train schedule, seat selection button, and payment confirmation page, simulating the user's finger movements step by step. For merchants who have already completed MCP or Skill integration, the AI can directly call standardized interfaces for a smoother experience; for the vast number of long-tail services that haven't yet been modified, screen reader provides the lowest-barrier compatibility path.

The problems with screen readers are also straightforward: stability hasn't been extensively verified. Mini-program interfaces vary greatly; dynamic loading, pop-up ads, and layout changes due to version updates all increase the probability of AI recognition failure. If a payment confirmation button is offset by a few pixels, can the AI accurately detect it? If misoperations occur during screen reading, such as misreading the amount or selecting the wrong delivery address, who is responsible? Alipay has not yet publicly disclosed relevant disclaimers or dispute resolution mechanisms.

The logic behind this approach is to first get users to use it. Once merchants see the order conversion rates brought about by AI, they will naturally take the initiative to connect to the standard interface to optimize the experience. This consumer-driven approach forces business-to-business (B2B) integration.

What do 300 million transactions prove?

Beyond its products and ecosystem, Alipay has done something else related to how AI makes payments.

At the AI Payment Ecosystem Conference in May 2026, Alipay disclosed that its AI payment transactions had exceeded 300 million, supporting 95% of general-purpose 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 addresses the issue of extremely small, high-frequency payments. When AI compares prices between two food delivery platforms, it might need to call a verification transaction of 0.01 yuan to confirm the account's validity; when AI selects the optimal combination from multiple coupons, each coupon verification constitutes a payment action. These transactions are small in amount, but far more frequent than those of human users. Traditional payment systems were designed for "human confirmation and human payment," but Token Pay delegates this action to the agent.

AI wallets are more like issuing a budget card to agents. Users set rules and limits, and the AI autonomously completes payments within those rules. Ant Group CEO Han Xinyi made a prediction at the conference: in the future, countless agents may be active in economic activities, and interactions will shift from human-to-human interactions to human-to-agent interactions, as well as interactions between agents.

While 300 million transactions may not seem like a large sum in terms of Alipay's total annual transaction volume, its significance lies in validating one thing: users are now allowing AI to fulfill real business transactions on their behalf, rather than just performing queries and price comparisons. From ordering a ride or food with a single sentence to AI-powered payment deductions, the technological and user authorization links for this service loop have been seamlessly integrated.

WeChat Pay has not yet publicly disclosed specific plans for AI-driven transformation. While WeChat Pay also covers a massive user base, its scenarios are more focused on social transfers, red envelopes, and merchant payment lock-in. The agent economy model may differ, and whether new differences will emerge in payment infrastructure between the two depends on whether WeChat AI Assistant will launch similar agent payment capabilities alongside its official release.

The ecosystem is being torn open in two ways.

Both Alipay and WeChat point to the Agent service entry point, but the different intermediate paths will tear two cracks that lead to different outcomes in the mini-program ecosystem.

Alipay's screen reader solution has passively enabled a large number of long-tail mini-programs to become AI-enabled. Merchants haven't done anything; users can already operate their services through AI. This will generate two reactions: some merchants will find that AI-driven order volume is increasing and will proactively integrate MCP or Skills to optimize the experience and secure more traffic distribution; others may resist because the source of orders has become ambiguous. Previously, every click a user made within a mini-program was traceable; now, merchants cannot obtain user behavior data for the portion of the path that AI screen readers operate on.

Alipay clearly anticipated this. LatePost reports that following the launch of the AI version of Alipay, an AI open platform for merchants and developers will also be released soon. This platform will likely address the question of how to allow merchants to benefit from the increased orders brought by AI while retaining visibility and control over service processes, user reach, and revenue distribution.

The pressure on WeChat is different. The high barrier to entry for source code licensing will divide developers into two groups. Top developers, with technical teams and commercial bargaining power, are willing to hand over their source code or invest resources in packaging skills in exchange for priority traffic distribution from WeChat AI Assistant. However, many small and medium-sized businesses may be unwilling to hand over their source code and unable to afford the packaging costs. If traffic does indeed shift towards authorized merchants after WeChat AI Assistant launches, unauthorized mini-programs will be marginalized in the AI service distribution channel. In the long run, WeChat's mini-program ecosystem may further concentrate on top players, which creates tension with WeChat's consistent emphasis on a "decentralized" ecosystem narrative.

A more subtle issue lies in technical standards. Alipay promotes its MCP (Multi-Channel Protocol), while WeChat defines its own set of Mini Program MCP protocols. Despite the similar names, their implementations are not entirely interoperable. A restaurant merchant wanting both Alipay AI and WeChat AI to access its ordering service might need to encapsulate its services according to both standards. This isn't a technical challenge, but it's a cost. Whichever side achieves economies of scale first will have greater bargaining power to push for de facto industry standards. With Alipay AI payments exceeding 300 million transactions, this advantage currently lies with Alipay.

Returning to the user side, the ultimate result of these changes could redefine the relationship between people and their phones. If Alipay's chat interface works, the frequency and scenarios in which users open Alipay will change. It won't just be opened when making payments, but rather when a need arises and a question is asked casually. If WeChat's Agent works, the way users interact in group chats will change. They won't need to leave 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 changed the question of which account users kept their money in. This time, the competition is about who users entrust with the task of "getting things done for me."

Twelve years ago, WeChat's red envelope feature was described by Jack Ma as a "Pearl Harbor attack." Twelve years later, after months of speculation surrounding WeChat's AI messages, Alipay has stepped into the spotlight. Which path truly reflects the real needs of the Agent era? The answer lies not in product launches, but in how millions of mini-programs are reactivated, and in the experience of hundreds of millions of users uttering "Help me" to their phones for the first time.

A whale bought 7,769 ETH in the past 3 hours, equivalent to approximately $17.46 million.

Popular Articles