# WeChat Tests Xiaowei AI Social Feature for Assistant-to-Assistant Chats > Source: > Published: 2026-09-08 07:24:29+00:00 **September 8, 2026**, (Inside AI) — **WeChat** is quietly testing a feature that lets two AI assistants talk to each other before their human owners get involved. The tool, called **Xiaowei AI Social**, is in a limited internal rollout. The feature lets a user's **Xiaowei** assistant contact a friend's Xiaowei assistant to handle an initial conversation. Possible tasks include arranging a meal, coordinating schedules, and checking preliminary interest. Users step in only for authorization and key decisions. This is not a public launch. The final release scope and availability have not been announced. The test signals a shift in how social platforms think about AI agents. ## Why WeChat Wants Agents to Talk First WeChat is the dominant messaging app in **China**, with over **1.3 billion** monthly active users. It is owned by **Tencent**. The company has been cautious about adding AI features that could disrupt user trust. Xiaowei AI Social flips the typical chatbot model. Instead of a human prompting an assistant, the assistants initiate contact. The human approves the outcome. This reduces back-and-forth messages for simple planning tasks. Tencent has not said how many users are in the test. The feature appears to target low-stakes coordination, not deep conversations. That is a deliberate design choice. It keeps humans in control of meaningful decisions. Industry watchers see this as a step toward **agentic AI** in consumer apps. Other companies are exploring similar ideas. **Meta** has tested AI personas in messaging. **Google** has shown agents that book appointments. WeChat's approach is more bilateral: two agents negotiate on behalf of two humans. One risk is user discomfort. People may not want a bot representing them in social contexts. WeChat's test is small, so the company can gauge reactions before a wider release. ## What This Means for Social Platforms If Xiaowei AI Social works, it could reduce message fatigue. Users would no longer need to send multiple texts to find a dinner time. The assistants would exchange availability and preferences first. But the feature also raises questions about authenticity. A friend might not know they are talking to a bot initially. WeChat has not detailed how the interaction is labeled. Transparency will be critical for adoption. Tencent has invested heavily in its **Hunyuan** large language model. Xiaowei is the assistant built on that model. The social test suggests Tencent wants to embed AI deeper into daily communication, not just as a standalone chatbot. The company faces competition from **ByteDance** and **Alibaba** in China's AI race. WeChat's massive user base gives it a unique testing ground. If the feature succeeds, it could set a template for AI-mediated social planning worldwide. No timeline has been given for a public release. The limited test will likely run for weeks or months. Tencent will watch how users respond to having their assistants speak on their behalf.