Zhihu Profits First Time, Eyes AI Dramas

🔑 Enhanced Key Takeaways

• •Zhihu's pivot toward AI-generated content (AIGC) leverages its proprietary 'Zhihu AI' large language model, which was specifically trained on its high-quality Q&A corpus to improve the efficiency of script-to-video conversion for short dramas.
• •The revenue decline is partially attributed to the strategic phasing out of low-margin advertising segments and the consolidation of its 'Salt' (Yanxuan) membership program, which now prioritizes high-value IP monetization over broad user acquisition.
• •Institutional investors have signaled caution regarding the sustainability of the profit, noting that the 2025 financial performance was heavily bolstered by non-recurring investment gains rather than organic growth in the core advertising or professional education business lines.

🛠️ Technical Deep Dive

• •Zhihu utilizes a proprietary multimodal pipeline that integrates its internal LLM for script adaptation with third-party video generation models (e.g., Sora-like architectures) to automate the production of 'AI short dramas'.
• •The platform employs a 'Content-to-IP' recommendation engine that identifies high-engagement narrative threads within its 16M-item tech/AI corpus to predict viral potential before greenlighting production.
• •Data infrastructure relies on a vector database architecture that maps user interaction patterns to specific narrative tropes, allowing for dynamic adjustments in AI-generated plotlines based on real-time viewer feedback.

🔮 Future ImplicationsAI analysis grounded in cited sources

⏳ Timeline