Wen/Chen Yongwei

On June 16, the AI version of Alipay officially launched and launched the invitation test, with an AI assistant named "A Bao" built in. Users only need to right slide into the AI version of Alipay on the home page of Alipay and give instructions to "A Bao". It can complete the tasks of taxi taking, registration, transfer, and searching for nearby idle charging posts, without having to look for related functions in the menus.

A few days later, WeChat also began grayscale testing its AI assistant "Xiao Wei". Users can not only send messages, make phone calls, post on WeChat Moments, adjust WeChat settings, but also use it to search for articles, generate images, and even call WeChat's internal mini programs to complete tasks such as registration, buying coffee, and presenting boarding passes.

Alipay and WeChat have launched AI functions one after another, which is easily reminiscent of their competition for Internet access over the past decade. This series of competition not only shapes the competitive pattern of China's Internet, but also promotes the process of national digitalization. Nowadays, the "war" between the two powers has resurged, but their competition for entry has shifted from mobile desktops and offline cash registers to AI agents driven by natural language.

So, compared with the competition for mobile Internet portals in the past, how is this AI portal war different? What value does it have in the context of global AI competition? What new regulatory issues will arise?

1、 Past War: Competition for Entrance in the Mobile Internet Era

At the beginning of the birth of the mobile payment track, Alipay once had an absolute advantage. Relying on the e-commerce transaction scenario precipitated by Taobao and Tmall, Alipay took the lead in building a complete account system, capital risk control and financial management matrix. In the fourth quarter of 2013, Alipay's share in the third-party mobile payment market reached 75.9%. At that time, WeChat was just a social app that focused on communicating with acquaintances, and its payment function was just an edge trial project. Few people believe that WeChat payment will soon become a payment giant on a par with Alipay.

The turning point occurred during the 2014 Spring Festival. On January 27, 2014, WeChat red envelopes were officially launched. Users only need to bind their bank card to participate in "grabbing red envelopes" and exchange red envelopes with friends.

Relying on the vast social network of WeChat, red envelopes are spreading rapidly among family and friends in a viral manner. From New Year's Eve to the eighth day of the lunar new year, over 8 million users participated in "grabbing red envelopes" and more than 40 million red envelopes were received. According to the financial industry insiders cited by the People's Daily Online Finance Channel, the increment of bank cards successfully bound through WeChat red envelopes should reach at least 30%. Later, Jack Ma referred to this incident as the 'surprise attack on Pearl Harbor'.

After years of reviewing this classic commercial raid, it is not difficult to see that WeChat invested heavily in launching the "red envelope war". In addition to seizing the market share of mobile payment, there is a more important strategic intention: to compete for the key entrance for people to enter the mobile Internet.

The so-called "entrance" refers to the first channel through which people enter the system of goods, services, and transactions. When users want to obtain news, search engines and information flow are the entry points; When wanting to connect with friends, social media is the gateway; When you want to purchase goods, e-commerce platforms are the entrance; When you want to complete offline consumption, payment tools are the entrance.

The importance of entrance is determined by the "platform" characteristics of the Internet economy. The Internet platform itself may not produce all content and goods, but it can gain huge economic power by controlling users' access to content, goods and services. In the case of limited user attention, whoever controls the entrance is more likely to decide what users see first, which merchants they enter, which services they use, and thus earn more commissions, advertising revenue, and user data.

Although payment itself is not the most profitable business, it is very important for the development of Internet platforms. Before the 'Red Envelope War', Alibaba firmly controlled the payment entrance, which created inherent advantages for its expansion into areas such as consumption, advertising, finance, and lifestyle services. For Tencent, which also wants to expand into these fields, acquiring payment entry points has become an important goal.

It is worth noting that in the "red envelope war", Alipay and WeChat reflected two different business logic. The logic of Alipay is "transaction first, payment later" - users first choose goods on Taobao, and then use Alipay to pay. Therefore, Alipay started from the end of the transaction, relying on its account, capital and credit capacity, and gradually expanded to finance, payment, transportation, medical care and government services, trying to turn the payment account into a life service center.

WeChat's approach is exactly the opposite. It first grasps the most frequent social relationships of users, and then embeds payments into chats and groups. Users bind their bank cards not because they specifically need a new payment tool, but to participate in social activities such as red envelope games. WeChat thus transforms social traffic into payment relationships, and then sinks from the connection between people to the connection between people and merchants, and between people and services. In short, Alipay expands from transaction to life, and WeChat penetrates from social interaction to transaction.

Since the "Red Packet War", the competition between Alipay and WeChat on entry has lasted for more than ten years, and the different logic behind the two super applications has also been constantly colliding.

The "scan code war" that broke out afterwards, superficially competing for what tools users use to complete offline payments, essentially competing for the digital entrance to offline commerce. WeChat's strategy is "high-frequency relationships drive low-frequency transactions", allowing users to complete payments and consumption without leaving social scenes. Therefore, the focus is on promoting "face-to-face payment", personal payment codes, and small and micro merchant payments.

Alipay's principle is "payment account aggregation life service". It focuses on funds, bills and credit, and constantly absorbs commercial and public services. Convenience stores, chain stores, restaurants and other standardized scenes have become the focus of its promotion of QR code.

The emergence of mini programs is a further upgrade of entrance competition. For WeChat, mini programs have given social portals the ability to host commercial services. Users do not need to leave WeChat, nor do they need to download and install a large number of independent applications, to complete ordering, hailing, shopping, booking and payment. WeChat has thus transformed from a chat tool to a lightweight service operating system. For Alipay, the applet is a tool to upgrade the payment account to a comprehensive life service platform. With the help of small programs, Alipay can centralize medical, transportation, government affairs, finance and local life services in the same account system, so that users can not only use Alipay at the end of the transaction, but also enter Alipay first when looking for and handling services.

Looking back at this history, we can see that WeChat and Alipay have fought in the fields of payment, QR code and applet, but these competitions have always centered on the entrance. Both parties have adopted vastly different business strategies, and these differences essentially stem from their respective product genes. In the era of AI, these characteristics remain the key to understanding the competition between the two.

2、 New AI landscape: AI entry strategies for two major platforms

The decade long entrance war has made WeChat and Alipay reach a balance: WeChat occupies the upstream of demand by virtue of social interaction and content, and Alipay controls the downstream of transaction performance by relying on payment, account and life services. For a considerable period of time, although the two sides have continued to engage in repeated confrontations in local life, financial services, and commercial traffic, this fundamental pattern has not undergone fundamental changes. With the advent of the AI era, the conditions for this equilibrium to hold are beginning to shake.

In the era of mobile Internet, the role of the platform is "service shelf" plus "traffic channel". Users need to provide clear requirements themselves, search, click, compare, and then enter specific services. In the era of AI, the role of platforms has gradually evolved into "demand agents" and "task dispatchers". Users only need to express one goal, and the platform's AI can understand the intention, decompose tasks, filter solutions, call tools, and complete transactions after obtaining authorization.

With the transformation of platform roles, the key to entrance competition has undergone a triple shift.

Firstly, the competition for entry has shifted from "where users enter from" to "who is the first to explain what users want to do". In the traditional mode, users have already identified their needs, such as taking a taxi, booking tickets, or paying fees. The key to platform competition is to let users open which app and click on which page. In the era of AI, users often only propose a vague goal, such as "help me arrange a weekend trip" or "how to spend less money recently". At this point, the power of intent interpretation becomes the key to the entrance dispute. Whoever can first obtain and explain the user's intention is more likely to occupy the upstream of subsequent service allocation.

Secondly, the competition for entry has shifted from "competing for exposure and clicks" to "competing for service selection and scheduling rights". In the past, although platforms could influence user choices through search results, homepage recommendations, and mini program ranking, users often still saw multiple merchants. In the era of AI, the intelligent experience provided by the platform directly filters and combines services in the background, and only provides one or a few solutions in the end. The importance of attention, which is traditionally considered a scarce resource, will relatively decrease, and whether it can enter the service call list of intelligent agents becomes the key to platform competition.

Once again, the entrance dispute has shifted from "the first step in controlling transactions" to "the complete chain of control tasks". The traditional entrance is mainly responsible for introducing users to a certain service, but subsequent searches, comparisons, and payments are still completed by users. AI agents will extend from understanding requirements to task execution and transaction settlement, providing users with one-stop services. Once the platform masters the entrance, it can influence not only where users start, but also the entire process of completing tasks for users.

These three transformations mean that the core assets of traditional portals are traffic, page location, and user duration, while the core assets of AI era portals have become model understanding ability, tool calling ability, account and payment permissions, real service ecology, and user trust. In the past, the focus of competition was' who has the most users', but in the future, the more important issue is' how many things users are willing to entrust to whom '. The original balance between WeChat and Alipay is difficult to maintain.

Although they also compete for AI entrance, Alipay and WeChat have different strategies, which are still determined by their product genes.

As an application born out of e-commerce and financial transaction ecology, Alipay's core background is strong transaction, strong account, strong finance, and strong tool attributes. For a long time, Alipay has been deeply engaged in high-value scenarios such as capital settlement, financial credit, government affairs and people's livelihood, and has connected a large number of internal and external services such as Gaode, Elema, Ant Wealth, and has a strong financial risk control and compliance system. This gives Alipay an advantage in professional service scenarios, but as a tool application, it also has shortcomings such as low opening frequency and insufficient fragmentation viscosity. Based on such ecological characteristics, Alipay's strategic logic is very clear: by restructuring the core interaction of applications, create a unified AI portal, and use service closed-loop to make up for traffic shortfalls.

Ant Group has a strong accumulation of AI technology. Thanks to the long-term accumulation of industry data and service experience, its self-developed Bailing big model can handle tasks in fields such as finance, government affairs, transportation, and local life. With relevant technical support, "Abao" can analyze users' vague and complex service needs and break them down into multi-level tasks.

These accumulations give Alipay advantages in interpreting user needs and connecting them with vertical scenarios. Alipay therefore chose "A Bao" as an independent first level page of AI Alipay, and used it to integrate its own capabilities within the platform to cooperate with third parties. After the user requests, "Abao" can perform service filtering, resource matching, and scenario scheduling, and complete relevant operations after user authorization. Since Alipay has connected a relatively complete scenario and service system, AI agents can connect multiple links in the application from demand understanding, service matching, order placement and performance, payment to after-sales query, forming a relatively complete one-stop process.

In contrast, WeChat, which relies on social ecology to grow, forms a sharp contrast with Alipay in AI portal design. WeChat has a large and high-frequency private social traffic, but its proprietary trading business is relatively limited. Its service supply relies heavily on third-party mini programs, and the platform as a whole emphasizes open connections with external merchants and service providers. At the same time, the demand generated within WeChat is characterized by fragmentation, small amounts, and high frequencies. Based on this ecological background, WeChat did not completely copy Alipay's model of taking independent AI pages as the center, but adopted the product logic of "independent entrance plus native contact embedding".

WeChat "microenterprises" do not rely only on a single model, but use both Tencent self research model and open source model. On the user interface, WeChat not only sets up an independent entrance for "small and micro" accounts, but also embeds AI capabilities into native touchpoints such as chat dialogue boxes, article reading, and mini programs, capturing fragmented needs through social dialogue scenarios and interpreting daily shopping, group buying, and travel demands. At present, WeChat's AI entrance focuses more on the lightweight requirements inherent in chat scenarios, turning social conversations themselves into carriers for capturing user intentions.

Due to WeChat's reliance on third-party ecosystems for transactions and service provision, many functions require support from external service providers. To this end, WeChat promotes the integration of third-party intelligent agents with mini programs through its AI development model and related interfaces. WeChat's AI agent can call from a service pool composed of third-party providers based on the parsed requirements.

These third-party programs come from different service providers. WeChat does not necessarily seek to close all complex tasks in a single application like Alipay, but attaches more importance to connecting different services through open scheduling. For tasks involving transactions, the payment process can still be handled by WeChat Pay.

Comparing the AI portals of the two, we can see that the evolution path of Alipay and WeChat agents is the extension of their respective ecological genes. Alipay adopts a centralized closed-loop approach, relying on vertical AI technology, rich service scenarios and financial level risk control system, trying to grasp the initiative of user demand interpretation, service scheduling and full process performance, focusing on high-value, long-term in-depth service scenarios; WeChat adopts a distributed and open approach, relying on high-frequency social touchpoints, third-party ecosystems, and payment bases to cover mass lightweight services and high-frequency daily consumption scenarios. The two paths are different, but their goals are highly consistent: to seize the core connection hub between users and digital services in the AI era, and complete the transformation from traditional traffic entry to intelligent task entry.

3、 Global Value: Application Innovation Path of AI Entrance

If you step out of the Chinese market and look at the construction of AI portals by Alipay and WeChat from the perspective of global AI competition, you will find that its significance may have gone beyond the entry competition itself.

For a long time, there has been a common saying in the field of digital economy: many technologies originated from European and American countries, but China has more advantages in using these technologies for combination innovation and creating new models. For example, the early product form of WeChat was inspired by instant messaging products such as KikMessenger － ger, and the early model of Alipay can also be found in overseas payment tools such as PayPal. However, the influence of KikMessenger has already declined significantly. PayPal's core business still focuses on payment and financial services, while WeChat and Alipay have grown into super applications. Similar combinatorial innovations seem to be repeating themselves in the field of AI.

In the past few years, the focus of the global AI industry has mainly been on basic models. AI companies compete with each other in terms of parameter size, reasoning ability, training cost, and benchmark test results, and the capital market is accustomed to using model capabilities to judge a company's competitive position. A model is like an ever smarter brain, capable of writing, programming, analyzing, and answering questions, but it still has a distance to go between it and the real economy.

This idea of overemphasizing the model and relatively neglecting the application makes it difficult to fully reflect the productivity effect of AI, and also intensifies the doubts of the society about AI foam. At the same time, the excessive pursuit of model capabilities and the disregard for commercial monetization capabilities have left many AI companies struggling to make ends meet. Even top companies like OpenAI and Anthropic find it difficult to achieve profitability. If this situation persists for a long time, it is obviously not conducive to the development of the entire industry.

As AI intelligent agent technology gradually matures, many leading enterprises are beginning to export AI capabilities to industry applications while exploring business models. OpenAI、An－thropic、 Google and other companies generally adopt a general model approach, attempting to use their own models as a starting point to output their capabilities to applications such as browsers, websites, office software, and business services. Apple, Microsoft, and Google rely more on operating systems, office software, and device permissions to embed AI agents into smartphones and computers. The basic route is to first build a sufficiently powerful intelligent core, and then gradually connect applications and services downwards. In short, it is about moving intelligence towards applications.

WeChat is different from Alipay. Their construction of AI entry points is more like adding an intelligent layer on top of the real service system. WeChat has connected people to people, people to content and people to applet, and Alipay has connected accounts, merchants, payment, finance and a large number of livelihood services. They do not need to build the mobile environment of intelligent agents from scratch, but only need AI to reorganize the existing digital infrastructure. In short, it is about embracing intelligence in applications.

Compared with the path starting from the model, this path has several advantages.

Firstly, it is easier to integrate AI with the real economy. In reality, service is not just a simple information Q&A, but a complete process consisting of identity, account, price, inventory, payment, fulfillment, and after-sales service. A universal model can recommend hospitals, but may not necessarily have real-time information sources or complete identity authentication; Can analyze consumption, but may not necessarily be able to access real bills. The advantage of WeChat and Alipay is that they are in these processes. Their AI does not search for scenarios from the outside, but directly embeds into existing transaction and service chains, making it easier to move from "being able to answer" to "being able to do things", and also easier to transform technical capabilities into actual productivity.

Secondly, it is easier to form a measurable and chargeable commercial closed loop. The value of many AI products mainly lies in abstract abilities such as better answers and faster writing, but it is still difficult to measure how much revenue these abilities have created. Service oriented intelligent agents can directly affect transaction rates, customer service costs, processing times, repurchase rates, and refund rates. As long as AI helps the platform facilitate one more transaction, reduces one manual service for merchants, or assists users in completing one consumption faster, its economic value can be directly observed and calculated. AI can also generate revenue through transaction commissions, merchant service fees, intelligent marketing, and value-added services, without relying solely on subscription fees.

Thirdly, it helps alleviate the contradiction of "high investment, low revenue" in the AI industry. Training and running large models require huge investments in computing power, chips, and energy, but ordinary users have limited willingness to pay for a single question and answer session. After WeChat and Alipay embed AI into payment, retail, travel, medical care and local life, the model cost can be shared by a large number of real economic activities. The business model of AI can also shift from "selling model capabilities" to "sharing actual results": no longer just charging by token or subscription, but receiving returns based on task completion, transaction facilitation, and cost savings. Users may not be willing to pay long-term for a smarter chatbot, but merchants are willing to pay for higher transaction rates and lower operating costs.

In this sense, the exploration of WeChat and Alipay on the AI portal may provide a more realistic and economic application path for Chinese AI enterprises, and provide experience for Chinese AI to find differentiated competition directions.

Of course, both WeChat and Alipay are backed by giants and have user scales, payment systems and ecological resources that are difficult for ordinary enterprises to copy. Their experience is difficult to be directly imitated. What other AI companies need to learn from is not to construct super applications like them, but to find the industry entry points they truly master. Manufacturing companies can start from equipment, processes, and supply chains, medical companies can start from diagnosis and treatment processes and patient management, and logistics companies can start from orders, warehousing, and distribution. Although these portals cannot be compared with the super portals of WeChat and Alipay, they can still provide sufficient industry data and application scenarios. As long as AI companies can establish a closed loop of "understanding requirements, calling tools, completing tasks, and generating profits" in a certain vertical scenario, and build AI capabilities on the entrance of the industry, they are also likely to achieve considerable economic benefits.

4、 Regulatory Puzzle: Multiple Governance Challenges for AI Entrances

WeChat and Alipay build AI portals to make services more intelligent. Although they can significantly reduce the cost of user services and transactions, they will also enable the platform to gain greater power than in the mobile Internet era. In the past, platforms were mainly responsible for displaying information and providing entry points, and most operations and decisions were still in the hands of users; Now, AI agents can directly assist users in completing transactions. As the platform's involvement in the trading process deepens, related issues and risks will also increase.

The first issue is the authorization boundary problem. To do things for users, AI agents need to read location, contacts, chat content, consumption bills, health information, and even financial assets. But the user's agreement to AI completing a certain task does not mean that they agree to the platform's unrestricted access to all relevant data. For example, if a user asks AI to arrange a family trip, AI may only need time, number of people, and budget, but the platform may also use this as an excuse to read more chat records and consumption history. For the sake of convenience, users cannot make very detailed authorizations every time they assign tasks, which may lead to disputes.

The second issue is the attribution of responsibility. In the past, the responsibility for transaction disputes could usually be relatively clearly divided between users, merchants, and platforms. After the introduction of AI agents, the transaction chain has become complex and opaque. A wrong booking may be due to the model misunderstanding the user's intention, the platform calling the wrong interface, or a third-party merchant providing inaccurate information, making it even more difficult to determine responsibility. If the platform claims that AI can automatically handle tasks for users on one hand, but at the same time shifts all responsibility to the "user's final confirmation" when errors occur, it will result in unequal rights and responsibilities.

The third is the conflict between commercial interests and user interests. AI agents are different from regular search in that they usually do not display a large number of options, but directly filter and even make choices for users. The platform thus gains stronger demand allocation power. If the platform prioritizes calling self operated services, affiliated merchants, or prioritizes products with higher commission payments, users may not realize that their choices have been influenced by commercial arrangements, and their interests may also be unknowingly damaged.

The fourth is the risk of financial security. AI agents can help users fill in information and compare plans, but the boundaries of automated execution must be particularly cautious when it comes to transfers, loans, wealth management, insurance, and large purchases. Once the model misunderstands user expressions or is affected by false information, malicious instructions, and interface attacks, users may suffer significant losses.

The fifth issue is the market dominance of super platforms. In the era of mobile Internet, the platform controls traffic through homepage recommendation, search sorting and external chain restriction. In the era of AI, this control may become more covert. The platform does not need to prohibit users from accessing competing services. It only needs to make its AI not call, call less, or prioritize calling these services, which can affect market competition. Small and medium-sized merchants may not even be able to directly reach consumers and can only strive to be included in the call list by the platform's intelligent agents.

In the face of these issues, regulators must also make corresponding adjustments, gradually establishing basic principles such as data invocation following the principle of minimum necessity, transparency in commercial recommendations, retention of manual confirmation for high-risk operations, full traceability of task execution, and fair access to third-party services. Only when regulatory measures keep up with the speed of industry development can the benefits brought by AI entry be fully released and related risks be effectively controlled.