从口语对话到习惯管理，从解题讲解到情绪支持，一类被称为「AI 学习伙伴（AI Learning Companion）」的新物种正在全球教育领域快速崛起。 本文将以技能训练、情绪陪伴、知识引导为核心维度，结合国内外产品演进，观察 AI 学习伙伴如何在具体场景中重写学生的学习方式。 01 AI语言陪练：当沉浸式对话不再依赖真人 语言学习一向是教育科技创新的敏感地带，也是「AI 学习伙伴」最早跑通的场景。在这个领域，高频开口、沉浸语境和即时纠错是提升能力的三大关 键。过去，这三项资源高度稀缺，往往依赖专业教师甚至母语者，一旦离开课堂，很难再维持相同强度。AI 在此填补了明显的供给缺口。 美国语言学习平台 Duolingo 是这一趋势的典型实践样品。为了弥补传统语言学习中「缺场景、缺陪练」的长期痛点，Duolingo 在 2023 年将 GPT-4 引入自 己的游戏化体系，推出了 Roleplay 对话功能。用户可以在咖啡馆、地铁、公司等虚拟情境中与 AI 自由对话，体验接近日常交流的练习氛围。与过去固定 脚本式的练习不同，AI 会根据用户的表达实时调整情节走向。如果用户提到「对坚果过敏」，扮演店员的 AI 便会立即 ...

Core Insights - The learning machine market in 2025 is characterized by innovation-driven market capture and the consolidation of advantages through continuous innovation [1] - The learning tablet market saw a total channel sales of 1.806 million units in Q3 this year, reflecting a year-on-year growth of 3.5% [2] - Education companies, represented by Xueersi, are expected to fully surpass traditional educational technology companies in 2024, capturing a dominant share of the learning machine market [3] - In 2025, education companies are not only focusing on providing quality content but are also deeply exploring AI teachers, which represent a higher barrier than quality content [4] Market Dynamics - Quality content serves as a strong barrier for education companies entering the learning machine market, with Xueersi's learning machine offering a total of 16 million minutes of courses, 4 million minutes added since launch, 850,000 sets of real questions, and 700 million questions in its knowledge resource library [5] - The brand reputation behind quality content is crucial, as using Xueersi's learning machine allows users to access lessons from well-regarded teachers with over 20 years of experience [6] AI Teacher Development - The shift from quality content to AI teachers is driven by the need for interaction and guidance in self-learning scenarios, as traditional human services cannot scale to meet the needs of millions of users [7] - AI teachers are seen as the most direct and effective form of deep service, leveraging vast amounts of user data generated during content usage to develop high-quality datasets for AI [8] - The evolution of AI teachers is likened to the progression of autonomous driving, with Xueersi's AI teacher currently at L3 capability, indicating a combination of multiple abilities [10][11] AI Teacher Capabilities - The L3 AI teacher represents a significant advancement, capable of breaking down complex problems into manageable steps and guiding students to think independently while providing real-time feedback [12][13] - The latest version, Xiao Si 3.0, offers a natural interaction experience akin to a real teacher, capable of immersive, multi-modal, and deeply personalized teaching [16][20] - The integration of multi-modal capabilities allows the AI teacher to interact with students in a paper-and-screen environment, providing real-time guidance and feedback as students work through problems [21][22] Competitive Landscape - The future competition in the learning machine market will center around AI capabilities, with successful products needing to provide both quality resources and effective solutions for utilizing those resources [22]

Core Viewpoint - The learning machine market in 2025 is characterized by innovation-driven market capture and the continuous strengthening of competitive advantages through technological advancements [1] Group 1: Market Dynamics - In Q3 of this year, the total sales volume of learning tablets reached 180.6 thousand units, reflecting a year-on-year growth of 3.5% [2] - Education companies, represented by Xueersi, are expected to fully surpass traditional educational technology firms by 2024, capturing a dominant market share in the learning machine sector [2] - High-quality content is identified as the core competitive advantage that allows educational companies to gain market traction [2] Group 2: Transition to AI Teachers - Educational companies are moving beyond merely providing quality content to deeply exploring AI teachers, which presents a higher barrier to entry [2][3] - The need for interactive guidance rather than one-way content delivery is a significant demand in self-learning scenarios [3] - AI teachers are seen as a solution to the limitations of human teachers, which cannot simultaneously achieve personalization, high quality, and scalability [3] Group 3: AI Teacher Development - The development of AI teachers is likened to the progression from L1 to L5 in autonomous driving, with Xueersi's AI teacher currently at L3 capability [4] - L2 AI teachers function as independent modules, while L3 AI teachers combine multiple capabilities to create a closed-loop teaching experience [7] - The latest version, Xiao Si 3.0, offers natural interaction and can provide personalized teaching experiences [9] Group 4: Technological Innovations - The integration of multi-modal capabilities allows AI teachers to interact with students in a more immersive manner, enhancing the learning experience [10] - The AI teacher can facilitate "paper-screen interaction," where students can write on paper while receiving real-time guidance from the AI [11] - The complexity of real-time judgment and interaction requires high-quality data and advanced R&D capabilities, which are essential for companies to develop effective AI solutions [13]