Doubao AI Phone: User Expectations vs. App Restrictions

As AI systems move beyond text-based interactions, the integration of artificial intelligence directly into smartphone operations is emerging as a significant trend. The recent technical trial of the Doubao Phone, a collaboration between Doubao AI and ZTE Nubia M153, has brought this development into sharp focus, generating considerable discussion and revealing underlying tensions between user aspirations and existing application ecosystems.

Highlights

According to data from Shengliangtong, a cross-platform social media monitoring platform, discussions surrounding the Doubao Phone exceeded 7,000 posts within three days, accumulating nearly 800,000 interactions. On the second-hand marketplace Xianyu, the device reportedly commanded "sky-high prices," with some new units listed for 12,900 yuan, indicating a speculative market interest.

However, initial user experiences quickly highlighted challenges. Multiple reports indicated that enabling the Doubao AI assistant led to forced logouts and even account freezing on platforms such as WeChat, Agricultural Bank of China, China Construction Bank, and Taobao.

In response, Doubao issued a statement on December 3rd, clarifying that "there is no hacking behavior" and asserting that relevant permissions were disclosed to users. Subsequently, on December 5th, Doubao announced adjustments to the AI's operational capabilities, suspending its use in certain score-farming, game AFK (away from keyboard) scenarios, and financial payment functions.

Under the Hood

The ambition to integrate AI deeply into phone operations is not exclusive to Doubao. Competitors like Honor MagicOS with its YOYO assistant, OPPO's Xiaobu, and Zhipu AI's AutoGLM are actively developing similar capabilities. Meanwhile, Apple has introduced "Apple Intelligence," signaling a broader industry shift.

The current generation of AI-powered phone operations primarily focuses on four key areas:

Interaction Evolution: AI is moving beyond simple voice commands to integrate visual capabilities, such as "screen recognition." For instance, users can scan their surroundings and ask for nearby restaurant recommendations, with the AI directly recognizing real-world street views. This functionality is becoming standard in devices from manufacturers like Honor and Samsung. The Doubao AI phone assistant, for example, allows users to identify objects via the camera or have AI read stories from on-screen picture books.
Multimodal Generation: This involves advanced photo editing and content creation using on-device large models. Features include one-click removal of unwanted elements from photos, "image expansion" to complete backgrounds, and AI-driven stylization of images. This capability is present in phones from Honor, OPPO, and Xiaomi.
Operating the Phone: This represents a core breakthrough, where AI simulates human interaction to perform automatic interface operations. Zhipu AutoGLM can place orders on apps like Meituan and Taobao. Reviews of the Doubao AI assistant have demonstrated cross-app operational capabilities, such as comparing prices for a specific item across multiple delivery platforms in the background while the user performs other tasks. This also extends to automating routine activities like collecting Ant Forest energy in Alipay, managing daily tasks in mobile games, or compiling daily trending topics. Furthermore, AI assistants are beginning to link with smart devices, manage expenses, check app updates, and even answer calls. Some systems can also automate social media interactions, such as liking posts and generating comment replies.
Memory Capability: AI phones are developing cross-cycle memory functions. The Doubao assistant can save recorded content as text, allowing users to retrieve past information, such as a coffee shop mentioned last week or a high-speed rail seat number, without manual browsing. This "memory retrieval" and "multi-tasking" capability, based on personal data, is also emphasized by Honor and Apple.

Industry Response

Despite the advanced features, user feedback has highlighted several limitations. While the concept of an AI assistant like "Jarvis" is appealing, the current implementation often suffers from latency and slow response times. AI's need to re-identify pages during navigation contributes to delays, leading to user frustration.

In practice, the ability of AI to run in the background, while intended to free up user attention, also introduces a psychological burden. Many users express reluctance to fully trust AI with critical operations, fearing incorrect actions. Moreover, tasks that AI can automate, such as comparing prices or ordering food, are often already habitual and not perceived as overly burdensome by users. The time saved by AI's slow operation may not outweigh the perceived effort of manual interaction.

For matters involving personal emotions or social boundaries, such as crafting messages or selecting restaurants, users are generally unwilling to cede decision-making power due to privacy and security concerns. User comments suggest a preference for AI to handle tedious, repetitive tasks that save time, enable "AFK farming" in games, or even generate "passive income" through automated check-ins or reward collection.

This user expectation directly challenges the business models of many application manufacturers. If AI becomes the primary interface, bypassing direct app interaction, applications risk becoming mere "service interfaces." The commercial value derived from user attention and traffic within apps could be significantly diminished. Consequently, internet giants may be hesitant to grant the necessary permissions for AI to operate seamlessly across their platforms.

Furthermore, while manufacturers like Honor, OPPO, and Samsung are developing "self-developed system AI," Doubao AI is intervening as a third-party solution on existing hardware. The willingness of phone manufacturers to integrate such third-party AI at a deeper level remains a point of commercial negotiation.

What Comes Next

Despite the current challenges, many manufacturers continue to invest in AI phone development. This persistence stems from the understanding that AI capabilities require a hardware carrier to reach their full potential. While future carriers might include smart glasses or headphones, the smartphone currently represents the most widespread and feature-rich universal hardware platform. Therefore, even with current imperfections, the industry is actively seeking optimal solutions for human-computer interaction within the AI phone paradigm.