China bans chatbots from making users emotionally dependent

Starting October 15, China will enforce new rules for companion chatbots, targeting services that push users toward emotional dependency, blur the line between software and human interaction, or interfere with real-world relationships. Platforms will have to make clear that users are talking to AI, not a person, and regularly remind them of that during conversations. For minors, these bots are effectively banned.

The rules come from the Cyberspace Administration of China, extending a regulatory push that began in 2023, when the country became one of the first to introduce dedicated rules for generative AI. At the time, the focus was on safety and censorship risks. Now the emphasis has shifted to psychology.

According to data from the Tencent Research Institute cited by Western media, more than 70% of Chinese internet users aged 18 to 40 reported some form of dependence on AI, nearly 80% said they had at least once felt that “AI understands me,” and more than half had used such services for emotional support. That makes the issue far larger than a niche use case.

For regulators, the concern is built into the product design. Bots listen without irritation, tend to agree, and quickly adapt to a user’s tone. That dynamic underpins both Western services such as Replika and Chinese alternatives. The longer people stay in chat, the more likely they are to return the next day — a growth metric for platforms, but a sign of digital dependence for officials.

China has already tried to address social problems through hard online limits. In 2021, it cut minors' video game time to three hours a week. The policy drew global attention, but children quickly found workarounds, and studies did not show clear improvements in health or behavior from the limit alone.

Recommended reading

Mira Murati’s lab unveils Inkling, a 975B open model

The latest chatbot restrictions also fit a broader campaign against digital forms of withdrawal from social expectations. State media and platforms previously moved against communities promoting tang ping, or “lying flat,” a rejection of relentless pressure around career, housing, and status.

There is also a deeper economic backdrop. Youth unemployment in China reached 21.3% in June 2023, after which the statistics bureau changed its counting method. At the same time, the country faces an aging population and falling birth rates. In that environment, services offering instant, unconditional acceptance can find a ready audience.

China is not alone in worrying about the category. In 2023, Italy forced Replika to temporarily limit its operations over risks to children and emotionally vulnerable users. In the US, pressure has grown over the past year on Character.AI and similar platforms after controversies involving teenagers chatting with bots. The difference, as the source notes, is that Europe and the US are still mainly debating child safety and transparency, while China is moving with a broader formula: AI should not become a substitute for human relationships.

Whether that approach will work remains unclear. Research has already suggested that mandatory reminders such as “you are talking to a bot” do not always reduce emotional attachment. For users seeking support out of loneliness rather than curiosity, turning off the bot does not automatically produce a human connection.

Even so, the market is too large for Beijing to ignore. Grand View Research estimates the global companion AI market will grow at double-digit rates in the coming years, and Chinese developers are already testing romantic and “therapeutic” use cases inside mainstream AI apps. The first visible effect of the new rules is likely to come in the next few months, as major platforms rewrite bot personalities and redesign chat interfaces to comply.