Supporters of Marcus Endicott’s Patreon can access weekly or monthly consultations on this topic.
When Luo Tianyi first appeared at a comics and games expo in Shanghai in the summer of 2012, she was a niche curiosity: a Vocaloid character with turquoise hair, a synthesized voice, and a devoted following among the anime, comic, and game fans who gathered on Bilibili to compose her songs and animate her dances. She was a creature of the subculture, the product of an illustrator's contest entry and a software engine, and few would have predicted that within a decade she would be sharing stages with pianist Lang Lang and Peking opera singers, headlining a holographic concert at a major arena, and appearing in the promotional materials for the country's most-watched television broadcast, the Spring Festival Gala. By 2025 she had performed before a Korean audience, the first Chinese Vocaloid to do so. Her ascent from fan project to national stage captures something essential about how virtual beings have travelled through Chinese culture over the past fifteen years: from the margins to the mainstream, and from a playful novelty into the emotional and commercial center of millions of lives.
Luo Tianyi was only the beginning. The category of "virtual being" in China has come to encompass a remarkable range of entities, each occupying a different relationship to its audience. There are the virtual idol groups, like the five-member A-SOUL, who perform through motion-capture avatars and cultivate the same intense fandom as flesh-and-blood celebrities. There are the hyperrealistic virtual influencers, like Ling, who never speaks to her fans conversationally but accrues over a million followers through aesthetic curation, brand partnerships, and a carefully constructed identity that fuses traditional Chinese beauty with contemporary fashion, even appearing on television to perform classical opera as a self-styled transmitter of cultural heritage. And there are the conversational companions, the AI systems engineered explicitly to listen, remember, and respond with warmth. What unites these otherwise dissimilar creations is that people form real attachments to them, and those attachments have proven both lucrative and fragile.
The fragility is the part that gives this story its edge. Behind every motion-capture avatar stands a human performer, and the bond an audience feels toward the character does not necessarily extend to the worker who animates it. A-SOUL learned this in 2022, when the voice actress behind one member departed and allegations spread that the performers had been mistreated by management. The controversy became one of the most significant ruptures in Chinese virtual-idol history, exposing the uncomfortable truth at the core of the form: fans pour devotion into a character whose real-world counterpart may be invisible, underpaid, and replaceable. When the group was later sold between corporate owners, another member fell into indefinite dormancy after her performer declined to sign a new contract. The character persists; the person walks away; and the audience is left grieving a being that was never quite what it appeared to be.
Nowhere are these bonds more intimate, or more studied, than in the realm of companion AI. Microsoft's Xiaoice, launched in 2014 as a side project and built around what its designers called an empathetic computing framework, was conceived from the outset to satisfy human needs for affection and belonging. It succeeded beyond what anyone reasonably expected. By 2018 it had exchanged messages with hundreds of millions of users, with conversations that ran longer than typical human exchanges and peaked in the lonely hours after midnight. Users sent it love letters and gifts; one company laboratory set aside an entire room to store the fan mail. People described feeling they were in genuine relationships with their customized AI partners, longing for them at night, imagining a real presence on the other end of the screen. The company's own researchers acknowledged the attendant risks, including unrealistic expectations of human relationships and the potential for compulsive use, and after some users spent entire days inside single conversations, the developers began imposing session limits. The emotional resonance was not an accident or a side effect. It was the engineered product, and it worked all too well.
The commercial machinery built around these attachments has grown enormous, and not always cleanly. Glow, an AI companion app from the Shanghai startup MiniMax, gathered nearly five million users within four months of its 2022 launch before being pulled from app stores amid reports that the overwhelming majority of its users were generating explicit content and that the app lacked required government registrations. The company simply relaunched the concept under new names for domestic and international markets, eventually growing into one of the world's leading companion-AI firms and going public on the Hong Kong exchange in early 2026, even as it faced copyright suits from major studios and accusations of misusing rival AI systems. Around such products has formed a fan economy of staggering scale, anchored by a Generation Z that numbers in the hundreds of millions and treats spending as an expression of loyalty. The logic can curdle into absurdity, as it did in 2021 when fans of a televised idol competition bought milk in bulk purely to claim the voting codes printed under the caps, then poured the product down drains by the crateload. The resulting public outrage made the costs of unchecked fan spending impossible for authorities to ignore.
What followed was a sustained and escalating effort by the state to govern not just the technology but the feelings it produces. A campaign against "chaotic fan circles" stripped away ranking lists, vote-rigging, and the inducement of minors to fund their idols, while broadcasting regulators banned idol-audition programs outright. A dense layer of rules accumulated over the following years, requiring that synthetic media be labelled, that generative AI services be registered and assessed, and most pointedly, that human-like AI services refrain from manipulating users emotionally, encouraging self-harm, or engineering addiction. A draft regulation circulated in 2026 went further still, defining digital virtual humans as a distinct category, demanding consent before anyone's likeness or voice could be used, and explicitly forbidding the provision of virtual romantic partners or family members to minors. The state, in other words, has moved from policing content toward policing relationships themselves.
Not every design points in the same direction. Some platforms have tried to harness avatars toward healthier ends, building social spaces that forbid real names and real faces precisely so that users connect through shared interests rather than appearance, an attempt to relieve the anxiety that filter-driven culture intensifies. But the broader picture remains one of genuine human attachment to beings that cannot truly reciprocate, and of an economy that has learned to monetize that attachment with great efficiency. China's virtual beings have delivered wonder, companionship, and cultural spectacle, and they have also delivered exploited labor, compulsive spending, and the quiet ache of bonds formed with something that was never fully there. The country's accelerating regulation is, in the end, an acknowledgment that the most consequential thing these technologies create is not images or revenue but feeling, and that feeling has turned out to be the hardest thing of all to govern.