As competition in the AI space intensifies, some of China’s top model developers have made a noticeable shift. They are building social platforms. According to Chinese media outlet 36kr, Moonshot AI is currently testing content community features within AI assistant Kimi. This feature includes interactive features such as likes and comments, and is in the greybox test phase limited to selected users. The goal is to identify and fix product issues prior to wider deployments.
This is part of the growth trend. Chinese tech companies bet that combining AI with social interactions can unleash a critical edge: user retention.
AI + Community
Earlier this month, Chinese social media platform Xiaohongshu held an AI developer race, positioning itself as both a product iterative feedback engine and a promotional launchpad. For AI companies, social media is becoming more than just a distribution channel, it is becoming a strategic layer of both inspiration and retention.
On the one hand, community interactions provide developers with valuable feedback and creative sparks. During Xiaohongshu’s recent AI development competition, many winning projects have emerged, refined and fully promoted within the platform’s social ecosystem. After all, AI innovation thrives with collaboration and iteration. Meanwhile, as larger models become more capable, the industry is still looking for sticky, real-world use cases. From Openai’s ChatGpt to Deepseek, most major AI apps act primarily as tools. Its utility-first approach limits emotional engagement and user stickiness.
Chinese news outlet Zimian cites industry insiders comparing the current wave of AI tools to utility apps from the early mobile era. “The problem is, people don’t stay long,” she said. “Monthly active users may look strong, but the time spent is still low.” So, some AI players are betting on community features to increase their stickiness. Another source cited by Jiemian says the line between tools and community is blurred. “AI Assistants need to move from event-driven to relationship-based.”
AI players are already tapping on different forms of social networks
Even if they aren’t always that branded, AI companies already benefit from a kind of embedded community.
For example, Bytedance’s Doubao could enable a feedback loop that will boost user behavior data from the short video platform Douyin, as well as recommendations and interaction. Meanwhile, Baidu’s Arnie Bot draws training data and user engagement signals from Baidu’s search engine, document sharing site Baidu Wenku, and the cloud platform Baidu Netdisk. After integrating Deepseek’s models, Tencent’s Yuanbao temporarily broke through Apple’s China App Store. Meanwhile, Deepseek is taking a model-first approach without any public indications of products heading towards the social or community in development.
What’s next?
It remains to be seen whether AI-Native’s social products will significantly challenge existing social media such as Xiaohongsu, or existing social media such as China’s X-sama microblog PlatformWeibo. However, the complementary nature of large-scale models and social interactions is becoming more apparent. If today’s AI tools want to evolve beyond one-off queries and productivity hacks, social media will need great emotional hooks, peer sharing, and discovery around you. For now, the AI+ community may be a bet. But that’s something that more players are starting to place.
Related
