A different AI entry point

While many large-model vendors are competing for a general-purpose AI gateway, NetEase Media is taking a more community-centered route. The company is positioning Beehive AI as an important consumer-facing AI entry point, with NetEase Xiaomifeng, a youth-oriented community product, serving as its main testing ground.

According to NetEase Media vice president Li Miao, NetEase has not just started investing in AI over the past one or two years. Over nearly five years, the company’s cumulative R&D spending has approached RMB 100 billion, with a significant portion going into AI-related research and capability building. Games were among the earliest areas of adoption: titles such as Justice, Eggy Party and Where Winds Meet have explored AI NPCs, AI editing assistants, and AI-generated scenes and content.

NetEase now wants to separate these AI capabilities from individual business lines and apply them to news, music and new consumer products.

Why Xiaomifeng matters

NetEase Xiaomifeng is described as an AI-native community for young users. It has accumulated ranking lists, encyclopedic entries, employment and academic information, and uses “Fengchao” to support group chats, clubs and interest-based interactions. Clubs from thousands of schools across China have joined the platform.

The company’s view is that AI should not require users to first learn prompt engineering. A prompt is the instruction a user gives to an AI system; better prompts often produce better outputs, but they also create friction. NetEase Media wants AI to work inside existing product contexts, using user behavior, interests and community data to reduce the need for manual prompting.

Key facts include:

• R&D base: NetEase’s cumulative R&D spending over nearly five years has approached RMB 100 billion.
• Early AI use cases: AI has already appeared in multiple games as NPCs, editing support and content-generation tools.
• Community assets: Xiaomifeng has built up campus knowledge, rankings, academic and job-related data, and school-club networks.
• Governance layer: The platform has more than 300 volunteer user administrators.

Simplicity, creativity and shared knowledge

Zhang Zhimin, vice president of NetEase Media and head of Beehive AI, summarized the product direction in three words: simplicity, creativity and knowledge co-construction.

Simplicity means lowering the barrier to using AI. During this year’s spring recruitment season, NetEase Media tested a job-hunting service that matched students with openings across the web based on their major, school, interests and career preferences, and also helped with resume submission. The idea was to let community data fill in context automatically, instead of forcing every user to become a “prompt engineer.”

Creativity is another area where NetEase Media sees differentiation. The company believes AI may do more than reduce the cost of producing images, text and video; it may create new interactive content formats. Creators can define how content changes based on user actions, allowing viewers to help determine the final experience. One Xiaomifeng user created an interactive fighting experience with AI. Zhang argued that the important shift is not the visual form, but the production threshold: a simple game once required programmers, artists and planners, while ordinary users may now attempt similar work through prompts.

Still, the technology is not yet at the “one sentence to generate everything” stage. In the case shown by NetEase Media, the creator revised the work about six times, checking the output and adjusting it each time. Zhang acknowledged that AI content creation remains relatively costly and that the platform still needs to make the process easier.

Knowledge co-construction may be the longer-term competitive layer. Zhang said that as algorithms become more similar, AI products may increasingly differ by the completeness, accuracy and uniqueness of their knowledge bases. Unlike traditional encyclopedias that focus on standardized records such as titles and resumes, Xiaomifeng’s campus entries may include teacher characteristics, club information and practical campus-life experience—knowledge that is hard to obtain externally. Users create and verify such information, which can then support AI answers or training.

Open questions: governance, privacy and business model

Embedding AI deeply into youth communities also creates challenges around authenticity, privacy, echo chambers and governance. Li said NetEase Xiaomifeng aims to filter content through community co-governance before providing it to AI systems, while pure AI-generated content will be labeled in line with national requirements. Zhang also stressed the need to distinguish between AI-generated and AI-assisted content, with governance helping users understand what is real and what is fictional.

On privacy, NetEase Media said users can decide through privacy settings which data may be stored. Data users do not want stored will not be retained, and data stored with consent will be encrypted. Even so, as AI becomes more personalized, avoiding stronger information bubbles will remain a continuing challenge.

NetEase is also introducing more Agent capabilities into the community. An Agent can be understood as an AI assistant that performs tasks over multiple steps. These capabilities are first offered to some highly active users to help with content creation and scheduling, and will later be integrated with Beehive AI. The “Longxia” assistant is expected to handle community tasks, summarize daily events, and eventually provide analysis and companionship.

NetEase has not set a clear commercialization timetable for Beehive AI. Zhang said the current priority is to verify what value the product can deliver. Future monetization may involve paid services or advertising, but only if users believe the ongoing value justifies the cost.

The strategic signal is clear: NetEase Media is not simply building another chatbot. It is trying to attach AI to community data, user relationships and content workflows. If its knowledge-building and governance mechanisms prove sustainable, community-native AI could become a differentiated path. The next test is whether young users will adopt it frequently in real scenarios without feeling extra friction.