By Ma Jinnan
Over the past couple of days, links for Yuanbao Party’s cash-red-envelope giveaway have blown up countless WeChat groups, leaving them scrambling for cash handouts.
From work chats and neighborhood groups to family groups and alumni chats, message after message saying "XX sent you a cash red envelope" spread like a virus and flooded everyone’s screens. Plenty of members were kicked out for breaking group rules by spamming what appeared ad messages. Some users were even said to have built an "anti-spam" setup—creating dedicated "AI red-envelope mutual-aid groups," treating the main groups as a "sandbox," and only allowing red-envelope links to be spammed inside the mutual-aid groups.
Some people joked that, judging by WeChat founder Zhang Xiaolong’s style, this Pinduoduo-like viral growth play wasn’t far from getting shut down by the WeChat ecosystem; others put it more bluntly: keep playing like this and WeChat itself is going to get wrecked.
So is Yuanbao Party’s screen-flooding red-envelope campartygn a grab for the key "entry point" in the AI era, or just another classic big-tech stress response—"marketing for marketing’s sake"?
Yuanbao Party’s sudden wave of spam-like virality may look abrupt, but it happened against a misaligned yet delicate backdrop.
ByteDance’s Volcano Engine was the exclusive AI cloud partner for the 2026 CCTV Spring Festival Gala, and ByteDance’s AI assistant "Doubao" served as the core interactive tool—deeply embedded into the live broadcast through interactive mechanics, and presented simultaneously across multiple terminals including Douyin, TV, and the Doubao app.
To secure the exclusive AI cloud slot for the Gala, ByteDance was bound to spend heavily. That meant that during the Spring Festival—the most important window for traffic and mindshare in China’s internet world—ByteDance had already seized the high ground of "AI + Spring Festival" ahead of everyone else. Tencent, of course, couldn’t just hand it over. But since going head-to-head to compete for the CCTV Gala slot wasn’t cost-effective, it chose a different route: rather than fighting ByteDance on the Gala stage, it launched a RMB-1-billion-scale red-envelope battle on its own relationship-chain home turf—using Yuanbao and "Yuanbao party" to pull off a "lightning war" before the Gala.
The problem is, this blitz looked far too rushed.
The Yuanbao product itself wasn’t built to absorb and channel the massive traffic brought in by red envelopes. What "heated up" wasn’t any feature or use case—it was just the marketing. Open the Yuanbao App and you’ll find "party" tucked away as a single entry in the bottom navigation bar. Tap in, and you land in a space that looks like a WeChat group, yet isn’t quite one: there’s @Yuanbao, image posting and meme-editing, and modes like "Watch Together" and "Listen Together," but the overall experience feels stiff and awkward, with a murky, hard-to-follow flow. Users are pulled in by the red envelopes, and their first reaction isn’t "What is this for?" but "How am I supposed to use this?" and "How is this ‘party’ any different from the WeChat groups I use every day?" From marketing to product to operations, there’s a hollow disconnect: the red envelopes bring people in, but once they arrive they’re dropped on the floor—no handoff, no guidance, and certainly no real social habit being formed.
This feels more like a big company making a "knee-jerk" decision at a special point in time: the boss says, "We have to make a big splash." Since the "red envelope" playbook was proven effective by WeChat, they approve a 1 billion yuan budget and replicate the old formula. Even if the product and operations teams sense something is off, they usually don’t dare say much; and when the red-envelope wave recedes and engagement can’t hold up, the ones who take the blame are still the product and ops teams.
WeChat’s Spring Festival red envelopes back then became an instant legend because red envelopes had sustained in-product usage scenarios: they were both a payment tool and a social lubricant; they were supported by a real relationship graph, and they were followed through by a whole set of downstream scenarios—offline payments, red-envelope transfers, tips, and more. Red envelopes weren’t isolated; they were the front end of the entire WeChat Pay system.
But Yuanbao itself is a large-model Q&A assistant, and it naturally doesn’t have a high-frequency use case like "red envelopes." The so-called "AI group chat" hasn’t become a must-have need either. If Yuanbao wants to promote a new kind of social tool—"Yuanbao party"—then, from a product-logic standpoint, it should first satisfy users’ core social needs: for example, why would people choose to chat or group-chat here instead of on WeChat, and what unique value can this provide that WeChat can’t? Only then should it layer on red envelopes as an added, value-enhancing mechanic—instead of putting the cart before the horse by using red envelopes to pull people in first and worrying about everything else later.
What’s more, rather than a social product, Yuanbao is better described as a community product—and community products don’t have an inherent genetic link to payments. Communities run on content, shared interests, and atmosphere; red envelopes can at most provide a short-term jolt, but they can’t create lasting stickiness.
QuestMobile data showed that by the end of 2025, Doubao’s weekly active users had reached 155 million, DeepSeek stood at about 81.56 million, and Yuanbao ranked third with 20.84 million—a sizable gap. Meanwhile, Alibaba’s Qwen surpassed 100 million monthly active users in a short period of time, and the battle for the AI "entry point" had already entered a white-hot stage.
Red-envelope spam across feeds can, of course, drive a short-term spike in downloads and opens. In this landscape, it’s understandable that Tencent was eager to fire a shot during the Chinese New Year window, but if it doesn’t address the fundamental issues at the product and scenario levels, even 1 billion yuan in red envelopes may buy nothing more than a crowd of "grab the red envelope and leave" passerby users. Once the red-envelope tide recedes, everything will return to square one.
Chinese New Year Season: Criticalfor Big Tech
Internet companies treating Chinese New Year as the "main battleground" for marketing has been an established practice, almost a tradition carried over from the "red-envelope wars." In 2015, WeChat used the Spring Festival Gala’s "Shake" feature to hand out red envelopes, catapulting WeChat Pay into the mainstream overnight and disrupting the Alipay-dominated mobile payments landscape. From that point on, Chinese New Year was no longer just a culturally significant holiday—it became a "super testbed" for internet products.
The reasons aren’t hard to understand.
After all, the Spring Festival is the highest-frequency, highest-density social scenario in Chinese society all year round. Almost everyone returns to their families and friend circles during these few days—there’s ample time, concentrated attention, and both emotional and monetary flows speed up. The Spring Festival is naturally suited to for viral dissemination——WeChat groups, Moments, New Year greeting videos, and blessing copy together form a massive social amplifier. As long as a mechanic is simple enough, it can spread across different circles in a very short time. What’s more, the Spring Festival is also a key moment for user habit migration: many users who usually don’t have the time or energy to try new products are more wiling in these few days to download a new app and try a new playbook.
That’s why the Spring Festival window has become a).replace(""", must-answer question" that no major player is willing to miss. Whoever tells the best story here is more likely to gain the upper hand in the coming year’s fight for user mindshare and ecosystem positioning. That’s also why this year, from ByteDance and Baidu to Tencent and Alibaba, all made AI products the protagonists during the Spring Festival—rather than simply competing for traffic in e-commerce, payments, or short video as before.
Tencent Yuanbao’s Stress-Response
On the one hand, Tencent is not in the lead in the fight for AI entry points.
Doubao, backed by Douyin and Volcano Engine, was far ahead in both user scale and ecosystem synergy; DeepSeek, relying on its technical reputation and the Spring Festival window, had once pulled off a round of explosive popularity; and Alibaba’s Qwen rapidly integrated Taobao, Alipay, Amap, and other Alibaba services, pushing its consumer-side user base past the 100-million level in a short time.
WeChat red packets, group-buy bargain games, and all kinds of viral growth driven by social gaming are, at their core, the same recipe. As the battle for AI entry points moves into the stage of competing for "mindshare and penetration," Tencent instinctively returned to the weapons it knows best: throwing money at red packets and engineering viral spread through its social graph. By tying red packets to Yuanbao, and then using "Yuanbao Party"—an AI group-chat gameplay mechanic—to funnel users into a new social setting, Tencent is essentially trying to replicate what it once did with WeChat Pay: the "red-packet ambush of Alipay." The difference is that this time, what it’s trying to "ambush" is the entry point to the entire AI era.
At a deeper level, Tencent is trying to use "Yuanbao party" to add a "social" piece to its AI narrative.
An AI assistant is inherently a one-on-one conversation, while group chats represent an extension of a social network and relationship graph. If Tencent can prove that "AI group chat" is a must-have, then Yuanbao is no longer just a tool—it can become a new hub that connects social relationships, content, and services. From Tencent’s perspective, this is even a way to hedge against ByteDance’s "AI assistant + short video + hardware" combo: you do "AI + content + hardware," and I’ll do "AI + social + red packets."
The problem is that this grand vision hasn’t been made clear at the product level or in users’ perception. Red packets do deliver explosive short-term distribution, but for most everyday users, Yuanbao party’s positioning, the boundaries of its use cases, and its irreplaceable value are still vague. It’s not that users don’t know what AI is—they don’t know "why I should chat in a Yuanbao group instead of in a WeChat group." If that question isn’t answered well, no matter how lively the red packets are, all you’ve done is bring people to an "empty plaza with no content."
Why Yuanbao Party Can’t Become the Entry Point of the AI Era?
To judge whether Yuanbao party can become the entry point of the AI era, we have to be honest and answer a few key questions from the standpoint of product design and user needs.
First, the prerequisite for "AI group chat" is that there’s a real "group" to begin with.
"AI group chat" sounds cool, but it isn’t a self-consistent scenario. It’s an add-on capability that rides on top of "group chat." Without "groups," there’s no such thing as )"AI group chat."
The underlying color of "groups" is social relationships.
Acquaintance groups, family groups, work groups, interest groups, temporary collaboration groups—every type of group is built on top of relationships that have already been accumulated on existing platforms like WeChat, QQ, DingTalk, and Feishu. It’s not that users don’t have "groups"; it’s that they already have too many. What users lack isn’t "yet another place where they can create a group," but "a reason worth moving my existing relationships over, or investing in new relationships for the long term."
As an independent platform, Yuanbaoparty is born missing the single most critical thing: a social graph.
Although it lets you invite friends from WeChat and QQ, this "pulling a group outside the body" model is, in essence, "yanking users out of the social space where they feel most comfortable and dropping them into a new, unfamiliar setting," and the migration cost is extremely high. It’s hard to expect someone to move their family group, classmates group, and work group to Yuanbaoparty just for "one more group where they can @AI." What’s more, Tencent’s own ecosystem already has social infrastructure like WeChat and QQ. If Yuanbaoparty doesn’t offer clear functional and experiential differentiation from them, it’ll be hard to shake off the awkwardness of being "redundant."
More realistically, big tech companies don’t necessarily succeed with every new product they build.
Over the past decade-plus, Tencent has certainly tried its hand at social and community products—Weishi, QQ Kandian, various community apps. Some even received massive resources, but in the end, the ones that truly established themselves on the main social battlefield were still WeChat and QQ.
Traffic transfusions don’t equal product success. What determines life or death has always been how well the product’s positioning matches user needs. Without a clear value proposition and a differentiated use case, even the strongest traffic is just a one-off "watering," and it can’t become a real "reservoir."
Second, you can’t migrate a networked social graph just by throwing money at subsidies.
Red-packet subsidies and cash incentives can work in certain scenarios, such as e-commerce, food delivery, and ride-hailing——these are point-to-point transaction scenarios, where users are the buyer and the platform or merchant is the seller. If the price is low enough and the service is good enough, users will switch. That kind of migration is linear and one-way, with a short, clear decision chain.
But social relationships are web-like, egalitarian, and long-term. If you want to move your entire social graph to a new platform, it’s not just you migrating—you’re also asking everyone in your network to migrate with you. Even if you’re willing, they may not be; and even if they are this time, it’s hard to sustain over the long run.
WeChat’s moat is deep not because it gives out the most "red packets," but because everything is entrenched there: your relationship graph, payment habits, the mini-program ecosystem, official account content, workplace communication, and more. The switching cost is comprehensive.
With red-packet subsidies, at most you can spur users to "try it out" in the short term, but it’s very hard to truly move social relationships. Historically, every product that tried to use subsidies to pry away WeChat’s relationship graph ultimately proved this. Users might occasionally open an app to claim a red packet, but real communication and interaction will still return to the place they’re most used to.
Third, how much of an experience upgrade can AI group chat really deliver?
Let’s break it down further and look at the real value of AI group chat across several typical scenarios.
First, familiar-circle socializing.
Group chats among people who know each other are essentially about emotional interaction and shared context. People gossip, talk about family, work, and trending topics—the goal is to "be seen by each other," not to "produce a precise answer." An AI agent’s role here is awkward: either it behaves like a chatterbox, constantly butting in and disrupting the emotional flow; or it acts like customer service, coldly serving up a "best answer"but that’s not what people need when chatting with friends. Friends don’t need AI as a "lubricant," nor do they need AI to feed real-time solutions. Most of the time, they just need a space where they can comfortably talk "nonsense." If you move this scenario offline, it’s even easier to understand: when you and a few buddies are drinking and shooting the breeze, wouldn’t it be awkward if a random passerby suddenly jumped in to join the conversation?
Next, stranger-to-stranger socializing.
Stranger socializing is a niche track in itself, and it’s highly utilitarian—either killing time out of boredom or driven by hormones. AI can certainly smooth out parts of the experience, like auto icebreakers and topic prompts, but it’s hard for it to become a "must-have." In stranger scenarios, what people truly care about is "the person on the other side," not "the process of chatting." Once AI intervenes too much, it starts to feel like "what’s the point of chatting with a robot?" Once the novelty wears off, it’s easy to slip into awkward small talk, and retention is naturally hard to guarantee.
Finally, let’s look at office settings and work group chats.
Work group chats look like the most reasonable and practical setting for AI group conversations. But once you look closely, you’ll find that AI is only an aid here, not a true necessity.
The core of a work group is delivering outcomes"——meetings exist to make decisions, discussions exist to push tasks forward, and all communication revolves around results. Tasks are broken down to specific people and tied to specific deadlines. AI can improve efficiency in one-on-one chats by organizing information, generating minutes, and reminding you of to-dos—but it doesn’t necessarily have to "jump into the group chat". Group chats are already overloaded with information; if an AI agent also keeps churning out summaries, suggestions, and reminders like a chatterbox, it may end up creating even more junk content.
The reason WeChat has kept rolling out Do Not Disturb for group messages, collapsing pinned chats" and similar features is precisely because it realized what users truly need isn’t more information, but "fewer interruptions". The smarter AI gets, the more it makes people look stupid —and that’s not a healthy direction for a product.
Seen through these scenarios, AI group chat is more of a "nice-to-have" feature than a "must-have" entry point. It can improve efficiency or entertainment in certain specific cases, but it’s hard for it to carry the future of an entire platform.
An Entry-point Product?
Looking back at more than two decades of changes on the internet, every shift in an "entry-point platform" was, at its core, driven by one of two variables: either the medium that carries information underwent a fundamental change, or the way people obtain information or solve problems was redefined.
In the PC era, the entry points were browsers and search engines, because it was a "Web-centric" world—users opened web pages in a browser with a keyboard and mouse, and found information through search.
In the mobile era, the entry points were operating systems and app stores, and—within the Chinese context—several gradually evolved super apps——because the smartphone became the new medium, apps became the new interaction unit, and users shifted from searching the web to opening an app.
The rise of short-video platforms, meanwhile, proved from another angle that an "entry point" can also be a window for efficiently accessing information and entertainment——an algorithm-recommended immersive feed has replaced part of the traditional search-and-browse behavior, after all, "laziness" is human nature.
From this, a few common patterns can be abstracted:
Either the medium that carries information changes (PC → mobile → potentially glasses, in-car terminals, robots, etc. in the future), and only hardware and software that fit the characteristics of the new medium have a chance to become the new entry point;
Or there is an efficient window for acquiring information (search, recommendations, feeds);
Or there is an efficient window for solving problems (service platforms such as payments, ride-hailing, e-commerce, food delivery, etc.);
Or there is an online mapping of real social relationships (social platforms, instant messaging).
In the AI era, the entry point most likely won’t be just a "chatty assistant." It is more likely to take one of the following forms:
A new hardware medium, such as AI glasses or AI robots, bringing entirely new modes of human-computer interaction and new paths to access information;
Or new OS-level AI capabilities——like a "system-level agent"——deeply embedded into all of a user’s digital behavior, becoming a true "digital butler";
Or a super app that not only has powerful AI capabilities but also fully encompasses social, payments, content, and service relationships——in China, the closest to this form is still WeChat.
Viewed from this angle, the Yuanbao camp at present is neither new hardware, nor a new OS-level AI, nor a super app that fully carries social and payment relationships——it is more like a "community with AI features," an attempt to build a "small living room" outside the existing social graph, but this small living room still lacks compelling reasons for people to stay for the long term.
Not Building yet Another Chat App
So if the Yuanbao camp can’t shoulder the responsibility, what should Tencent do to truly win the fight for the AI entry point? Rather than pouring all resources into red-envelope incentives to drive user acquisition for a standalone app, it would be better to turn the AI entry-point battle into "a battle for ecosystem power"—deeply embedding Hunyuan’s model capabilities into WeChat and WeCom, the two most important "relationship containers", and going one step further by moving down to the hardware layer to build a true AI phone.
Specifically, Tencent has three paths it could pursue in parallel, though with different priorities and sequencing:
Path one: System-level: build a "Yuanbao phone" and make AI truly an operating-system-level entry point.
If Tencent doesn’t want to be left behind at the hardware and AI operating-system layer by rivals like ByteDance, Huawei, and Xiaomi, it shouldn’t avoid the AI phone battlefield. On the contrary, it should embrace it more proactively.
The "AI phone"" here doesn’t mean simply preinstalling a standalone app. It means designing everything—from underlying architecture and chip selection to OS customization and AI capability integration—around how to make AI omnipresent like the system itself: invisible, seamless, and trustworthy.
Under this approach, Tencent could choose to work with manufacturers that have production capabilities but weaker ecosystem bargaining power—similar to the relationship between Huawei and Seres back then—and, through deep capital-level alignment, fully connect hardware and software end to end. On that basis, it shouldn’t push a separate "Hunyuan AI assistant" sitting on the home screen; instead, it should make "WeChat AI" the core entry point of the system-level assistant. The Hunyuan large model would serve as the underlying "brain" supporting the entire AI capability stack, while "WeChat AI" would be the most familiar and trusted "face" at the user-perception layer.
An AI phone like this could, whenever the user needs it, tap into WeChat’s social graph, payment capabilities, Mini Program ecosystem, and content services at the lowest possible cost. For example, you say: "Remind me to buy my wife a cake after work", and it won’t just set a reminder—it will also factor in your location and spending habits, and may even generate a store card and navigation for you directly inside WeChat. You say: "Send a 200-yuan red packet to colleague XX and wish him a happy birthday",and it can automatically draft a more personalized message and complete the payment immediately—everything happening inside a container the user already deeply trusts, rather than in a brand-new, unfamiliar app.
Compared with ByteDance Doubao’s AI-phone route of "barging into other apps", Tencent’s route is using WeChat as the central dispatch hub and, through Mini Programs and other controlled interfaces, enabling cross-app and cross-service invocation in a relatively gentle, secure way. This not only avoids direct permission clashes with other big tech apps, but also makes the broader ecosystem more willing to collaborate—because everyone is talking to a central scheduler, rather than handing control over to an outsider’s god hand.
The second path. Ecosystem-level: build Yuanbao into the address book, open up WeChat group permissions for "calling AI", and start testing the waters with WeCom.
If the AI-phone route is relatively long-term and complex, Tencent can quickly push a "compromise solution" within its existing ecosystem—one that neither fully remodels WeChat nor leaves Yuanbao drifting outside.
Embed Yuanbao into WeChat users’ address books, and allow users to call on Yuanbao in group chats to join discussions. In WeChat groups or WeCom groups, users could use a single entry point to feed a topic to Yuanbao, have it summarize, analyze, and generate a document inside Yuanbao, and then paste the results back into the group.
To prevent spam from bothering users, Yuanbao’s permission to participate in chats can be set extremely low—by default, it would be passively responsive, stepping in only when explicitly @mentioned or requested, and never proactively flooding the chat; then, based on user feedback and usage data, Tencent could gradually open up additional smart-assistance features.
An even more important next step is to test the waters in WeCom first. Enterprise users are more receptive to efficiency tools, and they’re also more willing to pay for real productivity gains; WeCom group chats are inherently more tool-oriented, so it feels more natural and more reasonable for AI to play the role of a "recorder, summarizer, and prompter" there than in a purely social setting.
This kind of trial strategy is "flexible on both offense and defense": if the experience is good, it can be rolled out gradually to WeChat’s entire user base; if the results fall short, it can remain confined to WeCom’s beta without affecting the core ecosystem.
That way, Tencent doesn’t have to force Yuanbao into becoming "another social tool," yet it can still let its AI capabilities create value within the existing relationship graph. Users don’t need to migrate their social connections just for AI; instead, AI quietly shows up in the scenarios they already use, in the form of "enhanced features."
A third path: organization and cadence—shifting from "marketing-driven" back to "product-driven."
Judging from how Yuanbao’s red-packet campartygn flooded people’s feeds this time, what Tencent truly needed to adjust wasn’t the budget, but the cadence and priorities.
Marketing can certainly be done—but only if the product can "hold up." You build the scenarios, features, and experience first, and only then talk about using red packets to drive new users, rather than burning money on red packets first and hoping product and operations can catch up later.
The battle for AI entry points is a marathon, not a "Spring Festival red-packet sprint." A 1-billion-yuan giveaway can of course buy short-term buzz, but if it can’t be converted into real retention and sustained usage, that spend is more like paying for "three days on the trending list" than buying "a ticket for the next decade."
Tencent’s core advantage has never been "who can burn more cash," but "who can manage social relationships and the services ecosystem better." Truly embedding Hunyuan’s capabilities into WeChat and WeCom—making AI an invisible foundational capability rather than a constantly scene-stealing "super player"—may be the approach that best fits Tencent’s DNA.
One genuinely interesting paradox of the AI era is this: when everything becomes "smart," the very concept of an "entry point" may be redefined.
In the past, when we talked about an "entry point," we usually meant a specific app, website, or hardware device—you opened it, and your world unfolded from there.
But in a truly AI-centered world, what users face will no longer be a set of isolated entry points, but an always-on, everywhere-present "intelligent environment." AI can be embedded in glasses, in cars, in speakers, in TVs, in mobile operating systems, and of course in super-apps like WeChat that carry countless social connections.
In that sense, what decides who wins may not be who builds yet another "chat app" or a "red-envelope virality" game, but who can more safely and credibly become the "underlying protocol" of users’ digital lives.
WeChat is already a kind of Chinese people’s "digital organ," and the role AI is most likely to play is precisely the layer of "nervous system" that makes that organ smarter and more responsive.
Tencent, of course, won’t easily hand that position to anyone else. But what it truly needs to do is not to build another "Yuanbao faction"; it is to make AI part of WeChat, part of WeCom, and even part of all its future hardware and services.
When AI is no longer an entry point you have to "remember to open," but rather a layer of capability you naturally reach when you use your phone and the internet, the debate over the "entry point" may no longer matter as much.
And those attempts to forcibly "grab the entry point" with red envelopes and screen-flooding stunts will ultimately be written into a footnote in the history of internet marketing—turned into a case study of "how a big tech company turned excitement into awkwardness overnight."
The Yuanbao Party is not up to the task of fulfilling Tencent’s AI ambitions.
