Doubao: Why AI Smartphones Are Dying on Arrival in China

Efforts to build a true "AI smartphone" in China have lately looked less like a technological breakthrough and more like a drawn-out war of attrition, in which the main enemies are not a lack of computing power but the owners of app stores. ByteDance's Doubao assistant is a perfect example. When the project launched, it seemed we would finally see a union of a powerful language model and a mobile interface.

Reality turned out to be far more prosaic: after just one month of active use, the service faced heavy pressure from the tech giants that control the hardware market. In the end, the ambitious project had to resort to a kind of "self-castration" simply to keep its right to exist inside its competitors' ecosystems. The problem is that in China, the idea of an AI phone today has very little to do with artificial intelligence itself.

It is primarily a question of control over the user interface. When ByteDance deploys Doubao, which can manage system functions, answer calls or intercept notifications, it stops being a merely useful utility. It becomes a "Trojan horse" that takes power away from hardware makers such as Huawei, Xiaomi or Oppo.

These companies have spent years building their walled gardens, and they are not at all ready to hand the keys to the gate to an outside player, even if that player created TikTok.

The Doubao situation recalls the historical anecdote about Empress Cixi's first train ride. She was reportedly horrified by the steam and noise, seeing in the new technology a threat to the established order. Modern Chinese vendors behave the same way. In deeply integrated AI they see not progress but a threat to their role as middlemen. If users talk only to Doubao, why would they need the phone maker's custom skin? Why would they need the built-in app store or the vendor's promotional push notifications? In this scenario the AI assistant turns an expensive branded smartphone into a "dumb pipe" for data, stripping the manufacturer of its main asset: direct access to the consumer's attention.

As a result, ByteDance had to make humiliating concessions. The assistant's most interesting features, the ones that required deep system access, were cut. Now it is just another chat app in a pretty wrapper, little different from hundreds of others. This defeat is strategic, not technical. It shows plainly that under current market conditions no "revolutionary" software can beat proprietary hardware. You can train a model on trillions of tokens, but if the "Home" button belongs to your competitor, your model will never become truly smart.

This case matters for understanding where the industry is heading globally. We can see the beginnings of the same conflict between Apple and OpenAI, though there the sides are still trying to reach a deal. In China, the masks are off: either you make the phone yourself, or your AI lives on a reservation, tightly bound by permissions and security policies. For ByteDance, this means it must either buy its own smartphone maker or accept the role of content supplier, and not the operating system of the future. For now, a "smart" smartphone in China is above all an obedient smartphone, one that does not get in the way of its makers earning money from advertising.

The big question: Will ByteDance buy what remains of Smartisan or another brand to release a truly free AI phone, or has the era of independent assistants ended before it began?