YuanbaoPai, Wenxin, UC... Why are Baidu, Alibaba, and Tencent all rushing to build AI communities?

< p > Recently, China's former internet giants, "BAT," have rarely responded almost simultaneously to the product form of "AI group chat" within the same time frame: Tencent launched "YuanbaoPai"; Baidu's Wenxin is conducting internal tests on multi - character group chats; UC Browser under Alibaba has launched multi - agent group chats... < p > In the history of internet products, this kind of "tacit understanding" usually means that the giants have smelled the scent of a certain paradigm shift, and it also forces the industry to pay attention to the seemingly ordinary product form of "AI group chat." < p > The trend of "AI group chat" blew in from across the ocean. < p > In December 2024, ChatGPT began testing the group chat function in some regions and officially launched it globally on November 20, 2025. Soon after, Microsoft and domestic internet giants successively accelerated the promotion of group chat functions with multi - agent and multi - character collaboration. < p > Group chat, originally a basic function of social products, has suddenly been placed at the core of the AI era. This is not accidental. Technically, large models already have the basic ability for multi - round understanding and complex collaboration; in the market, the narrative of pure Chatbots is gradually losing its appeal, and multi - Agent collaboration has become a new story. < p > This points to a fundamental question: < strong > In the AI era, what form will the "super entrance" take? < p > The battle for entrances in the mobile internet era revolved around search, browsers, social media, and information streams. Today, against the backdrop of the convergence of large - model capabilities and the increasing homogenization of single Chatbot products, AI group chat may be a crucial move by big companies for the "AI super entrance." < h2 > < strong > 1. Beyond "AI Socializing" < p > Strictly speaking, OpenAI is not the first company to test the waters of "AI group chat." Some relatively niche domestic products explored this direction earlier. < p > In March 2024, "Laolao," an AI companion app under Shunwang Technology, launched a group chat function. Users can select from thousands of AI characters to create multi - person chat scenarios. Different AI characters participate in conversations according to their personas, simulating an experience close to a real - person group chat. This is more like an AI social experiment, focusing on the "sense of companionship," "immersion," and emotional value. < p > Subsequently, QQ tested "QQ Intelligent Agents," and its product form is closer to Minimax's overseas AI social product, Talkie. Users select characters in the intelligent agent square, and some characters are obtained through a card - drawing mechanism, emphasizing character scarcity and interaction stickiness. < p > According to the Top 100 GenAI Application Report previously released by a16z, similar applications centered around role - playing have much higher user retention and active time than ordinary tool - based products. < p > At this stage, AI group chat is essentially an extension of "AI socializing," emphasizing companionship, role - playing, and emotional connection. Its core goal is not efficient collaboration but to enhance immersion and interaction stickiness. < p > The real turning point comes from another path defined by OpenAI: group chat. < p > In this mode, real - life users collaborate with one or more AIs in the same group chat, and the AIs have a certain degree of initiative, memory ability, and context - understanding ability. This constitutes a task - oriented "multi - person + multi - Agent" collaboration mode. < p > The "tasks" mentioned here are not limited to serious work tasks but also include activities in daily life such as shopping, watching movies, chatting, travel planning, information retrieval, and content creation. These can be incorporated into the same "task flow" and completed on the same interaction interface. < p > This is why, in the products of giants, AI group chat is more often embedded in office, tool, and service scenarios rather than simple social entertainment. < p > Take Tencent's newly launched YuanbaoPai as an example. In a way, its form is more like a combination of "WeChat + Tencent Meeting + AI assistant": Users can have casual chats, share screens to watch videos together; they can also hold remote meetings and conduct teaching. At the same time, the AI can play dual roles as a task assistant and an emotional companion. Ideally, users can obtain both emotional and functional value in YuanbaoPai. < p > Similarly, the multi - AI character group chat in the internal test of Baidu's Wenxin App also emphasizes scenarios such as workplace collaboration and family life coordination. The AI can understand the context and actively participate in discussions. Baidu has also made it clear that the company is not trying to create a replacement for WeChat but is focusing on task requirements and using AI capabilities to help users achieve specific goals. < p > In comparison, Alibaba's actions in AI group chat are a bit more conservative. It did not directly open the "AI group chat" function on the recently popular Qianwen but quietly built in AI group chat in the now less - prominent UC Browser. < p > The AI group chat function in UC Browser uses a multi - Agent architecture. The default members include "Xiaoyou," the native AI assistant of UC Browser, Quark AI, Tongyi Qianwen, and the externally - cooperated DeepSeek. Users can ask the same question to multiple intelligent agents in the group at the same time, and different Agents will give independent replies based on their own characteristics. < p > However, different from Tencent and Baidu, the group chat in UC Browser cannot integrate real - person group chats and does not have social attributes. < p > From these product forms, although it seems to be "group chat" on the surface, in essence, it is gathering the AI capabilities originally scattered in different tools into a unified collaboration space. < p class="image - wrapper">< img data - img - size - val="1080,810" src="https://img.36krcdn.com/hsossms/20260128/v2_313499eb9f91442eb7a241a52f44adee@5091053_oswg122471oswg1080oswg810_img_000?x - oss - process=image/format,jpg/interlace,1"> < p class="img - desc"> Left: Yuanbao app; Middle: Wenxin app; Right: Screenshot of UC from Jazi Guangnian < h2 > < strong > 2. What Can "Group Chat" Change? < p > Looking back at the historical experience of the mobile internet: The reason why WeChat was able to defeat QQ was not due to disruptive technology but rather "minor differences" in user experience. < p > One of the key differences is that QQ emphasizes "online/offline." Users usually log in only when they need to initiate a conversation, find a specific person, and talk about a specific thing. In contrast, WeChat weakens the real - time status, allowing users to be "disturbed at any time and end the conversation at any time." This minor difference ultimately shaped the "always - online" social experience. < p > For the platform, "always - online" means that users spend more time, maintain more relationships, and engage in more behaviors in the same entrance. Therefore, it goes beyond the tool attributes of communication apps like MSN and becomes a daily state. The platform can build a large - scale traffic pool based on this and thus derive a series of monetization opportunities. < p > This is the foundation for the success of super - apps in the mobile internet era and also the core bottleneck for the long - awaited appearance of a super - app in the AI era. < p > A series of single - person dialogue products in the form of Chatbots, such as ChatGPT, Gemini, Qianwen, and Doubao, can hardly naturally form this "always - online" usage state. It is more like a super - tool that users use and then leave, rather than a virtual space where social relationships and collaborations can be precipitated. < p > AI group chat fills this gap. < p > Taking YuanbaoPai as an example, after in - depth experience by "Jazi Guangnian," it was found that its core difference lies in embedding AI into all possible social relationships of users. < p > Whether it is a private space like a family group or a friends group, or a public or semi - public space like a work group or a project group, AI can always be present as a general member, seamlessly integrating into the user's social relationship network and allowing information, knowledge, and even services such as shopping and taking a taxi to flow naturally in the interaction between humans and AI. < p > At the same time, the social link between people is shortened again. < p > Traditional social software is more based on acquaintance relationships. However, in real - life work and life, there are a large number of collaborative relationships based on weak connections. People do not establish long - term social relationships but only aim to complete a certain task or convey a piece of information. < p > In this case, the friend - adding mechanism of "application - verification - approval" seems cumbersome. In products like YuanbaoPai, members pulled into the group chat can directly initiate conversations with each other, eliminating the friend - adding process. This makes weak - connection collaboration in the real world more efficient in the digital space. < p > This means that users no longer need to establish a long - term relationship with someone just to ask about something. Instead, they can focus on specific matters, and the relationship will naturally fade out after the matter is finished. This model is closer to the way of collaboration in the real world. < p > From a more macroscopic perspective, this also corresponds to a shift in the interaction paradigm. < p > In the past, people interacted through GUI, and information and functions were fragmented into countless App islands. Now, ideally, AI group chat can integrate the capabilities of apps such as WeChat, Feishu, and Tencent Meeting into the same workflow. Users only need to dispatch multiple intelligent agents through natural language, and the AI will autonomously call tools, break down tasks, and execute them. < p > Based on this, AI group chat not only changes the social form but also provides a new human - machine interaction paradigm and may become the prototype of the AI entrance. < h2 > < strong > 3. The Real "Battle for Entrances" Finally Arrives < p > In the longer term, the collective layout of AI group chat by giants is essentially an advanced version of the battle for AI entrances. < p > When AI group chat becomes the core collaboration space for work and life, users' relationships, communication, decision - making, execution, and even creative behaviors are accumulated here. This "group chat" is no longer just a function but becomes the primary site for daily activities. The user stickiness and data value it brings will increase exponentially. < p > When users' relationships, project data, and collaboration history are all accumulated on a certain platform, the cost for users to migrate will also rise sharply, forming a network effect similar to that of WeChat and constituting the real moat for the super entrance in the AI era. < p > At the same time, AI group chat is also a natural interaction interface for multi - Agent systems. Functions such as e - commerce, payment, content, and enterprise services can all be embedded as Agents to form a new business closed - loop. < p > For example, in an ideal state, within a project group, AI can directly call design tools, order services, generate contracts, and complete payment settlements. Socializing, communication, tool use, and business transactions are completed in the same context, with a very short closed - loop. Alibaba has been trying to build such a capability system through Qianwen, but it has not yet launched the group chat function. < p > In this process, the boundaries of the "platform" formed in the internet era may be re - defined. < p > A successful AI entrance is no longer just a single super - app but a collection spanning multiple apps: calling various AI capabilities and tools at the bottom and hosting real - world social and business activities at the top. In the future, theoretically, users only need to pull different people and create different groups based on the same entrance, and AI will participate in a more "invisible" way. < p > Therefore, if multi - Agent and multi - person interaction are truly popularized through an entrance, then whoever controls this entrance is more likely to form a stronger monopoly effect in the AI era than in the mobile internet era. < p > From historical experience, every change in the form of the entrance ultimately changes not only the traffic distribution but also the operation mode of the entire industry and society. Browsers defined the entrance in the PC era, and WeChat defined the entrance in the mobile internet era. AI group chat is likely to become the prototype of the entrance in the AI era. It not only determines who has the users but also determines who will master the dominant power of the game rules in the future human - machine collaboration network. < p > Of course, the "AI group chat" of each company is still in the trial - and - error stage. The giants are approaching from the perspectives they are most familiar with: Tencent from the perspective of socializing, Alibaba from the perspective of ecological integration, and Microsoft from the perspective of office work. Whether a single super - entrance or a multi - entrance co - existing structure will be formed remains to be verified by time. < p > But it is certain that the long - awaited battle for entrances, which has been talked about for three years, may really break out in 2026. < p class="editor - note"> This article is from the WeChat official account < a target="_blank" rel="noopener noreferrer" href="https://mp.weixin.qq.com/s/i - AYfN9XPO3jHqYhviGH5A">"Jazi Guangnian." Author: Liu Yangnan, Editor: Li Zi. Republished by 36Kr with permission.