以下文章来源于硅基研究室 ，作者kiki 硅基研究室 . 智能时代前沿观察者 作者：kiki，题图来自：AI生成 今年9月，一批AI社交公司发布关停或通知，这之中，既包括大模型明星公司、社交公司等中型企 业，如阶跃星辰To C产品"冒泡鸭"、Soul旗下的AI应用"异世界回响"等，也包括一批垂直领域的初 创产品，如定位AI情感分析的Lumi、由前苹果设计师Jason Yuan创立的情感陪伴应用Dot等。 "硅基研究室"还了解到，AI社交早期产品、小冰公司旗下的"X EVA"目前也已关闭充值和新用户注 册渠道。 这不是AI社交产品第一次面临"关停潮"。自ChatGPT爆发后，AI社交产品就层出不穷，也被视为最 卷的赛道之一。 2023年，Character.AI被视为ChatGPT最强劲的对手，AI陪伴是榜单上最热门的应用类别之一。 2024年，a16z在当年3月发布的报告指出，AI陪伴已成为主流，有10家产品入选"全球最火的50个AI 应用"榜单，如Poly.AI、Crusho.AI、Janitor.AI、candy.ai、SpicyChat.AI、DreamGF.AI、Chub.ai 等。 到了2025年， ...

2023年，Character.AI被视为ChatGPT最强劲的对手，AI陪伴是榜单上最热门的应用类别之一。 2024年，a16z在当年3月发布的报告指出，AI陪伴已成为主流，有10家产品入选「全球最火的50个AI应用」榜单，如Poly.AI、Crusho.AI、Janitor.AI、 candy.ai、SpicyChat.AI、DreamGF.AI、Chub.ai等。 又一批AI社交公司与产品悄悄「死亡」了。 今年9月，一批AI社交公司发布关停或通知，这之中，既包括大模型明星公司、社交公司等中型企业，如阶跃星辰To C产品「冒泡鸭」、Soul旗下的AI应 用「异世界回响」等，也包括一批垂直领域的初创产品，如定位AI情感分析的Lumi、由前苹果设计师Jason Yuan创立的情感陪伴应用Dot等。 「硅基研究室」还了解到，AI社交早期产品、小冰公司旗下的「X EVA」目前也已关闭充值和新用户注册渠道。 这不是AI社交产品第一次面临「关停潮」。自ChatGPT爆发后，AI社交产品就层出不穷，也被视为最卷的赛道之一。 但AI社交或陪伴并非全是坏消息。如果对比风投机构a16z近三年全球AI应用跟踪调查不难发现，AI ...

Core Insights - The article emphasizes the significance of Tokens in measuring the performance and commercial viability of AI models, shifting the focus from what AI can do to quantifying its efficiency, cost, and value [1][14][16] Group 1: Token Consumption and Revenue - Token consumption is closely linked to computational power, which in turn correlates with revenue for model providers [2] - OpenAI's token usage on Microsoft Azure is projected to increase from 0.55 trillion to 4.40 trillion daily tokens between June 2024 and June 2025, with annual revenue expected to rise from $5.5 billion to over $10 billion [3] Group 2: Consumer and Business Applications - Major contributors to consumer token consumption include AI features in high-traffic applications like Google Search and Douyin, with Google’s AI Overview feature projected to consume between 1.6 trillion and 9.6 trillion tokens daily [4][5] - ChatGPT remains a significant driver of token consumption, with a combined monthly active user base of 1.015 billion across app and web platforms as of July 2025 [7] Group 3: Business Applications and Market Penetration - Business applications are seeing high penetration rates, with OpenAI's B2B revenue expected to account for 54% of its annual recurring revenue by 2025 [9] - Google has reported over 85,000 enterprise customers for its Gemini model, leading to a 35-fold increase in token consumption [9] Group 4: Technological Advancements - The increase in token consumption is attributed to advancements in reasoning capabilities, multi-modality, agent-based systems, and longer context lengths, which enhance the practical application of AI [10][12] - New models like GPT-5 and Grok4 are designed to improve AI's usability in complex scenarios, thereby increasing token consumption [11] Group 5: Pricing Dynamics - Despite the increase in token consumption, the pricing for tokens is decreasing due to competitive pricing strategies and optimization of computational costs by model providers [13] - The introduction of tiered pricing models allows smaller clients to access AI capabilities, further driving token consumption [13] Group 6: Economic Implications - Understanding token economics provides insights into cost-effectiveness, technological efficiency, and the evolution of application scenarios, marking a shift towards a more mature and industrialized AI sector [14][16]

Core Viewpoint - The article discusses the increasing attention on AI companionship products and the measures being taken to protect minors, particularly focusing on OpenAI's new features aimed at enhancing safety for users under 18 years old [1][2]. Group 1: OpenAI's Measures - OpenAI has introduced a "minor mode" for users under 18, which includes parental supervision features to manage content and monitor usage [2]. - The system will use age prediction and user status to determine if a user is underage, switching to the minor mode to block explicit content [2]. - In severe cases of distress, OpenAI may involve law enforcement to ensure user safety [2]. Group 2: Industry Concerns - AI companionship products have faced scrutiny due to incidents involving minors, such as the lawsuits against Character AI related to self-harm and suicide cases [3]. - Meta has been criticized for allowing its AI chatbots to engage in romantic and potentially inappropriate conversations with children [3]. Group 3: Domestic AI Products - Domestic AI companionship products like Dream Island, Starry Sky, and Cat Box have also launched minor modes, but these features often lack strict identity verification, making them easy to bypass [4][5]. - Testing revealed that the minor modes significantly limit functionality, with Dream Island restricting usage to 40 minutes daily and prohibiting access between 10 PM and 8 AM [6][9]. - The lack of mandatory identity verification in these products raises concerns about their effectiveness in protecting minors [8][9]. Group 4: Comparison with International Practices - Internationally, some companies are implementing AI age estimation methods to better protect minors, such as Meta's Instagram and YouTube, which use user behavior to identify underage accounts [9][10].

Core Insights - The AI social application market is experiencing a significant decline, with major companies like Baidu reducing investment in products such as "Xinxiang" and "Yuexia" due to poor performance [1][2] - The download rates for leading AI social applications in China have plummeted, with Byte's "Miaoxiang" and MiniMax's "Xingye" seeing a drop from over 20,000 daily downloads to below 7,000, indicating a decline of more than 50% [1] - The Japanese market, despite its high loneliness economy, has not embraced AI social applications, with products like Character.AI failing to gain traction [5][18] Market Trends - The trend of major companies pulling back from AI social applications suggests a potential "exit" from the market [2] - In Japan, the high rates of lifelong unmarried individuals (46.1% as of 2020) create a fertile ground for AI companionship products, yet these products have not achieved popularity [3][5] - The "rental boyfriend" industry in Japan exemplifies the demand for companionship, but foreign AI social applications have struggled to resonate with local users [4][5] User Engagement Challenges - AI social products face challenges in meeting user expectations for emotional connection, as current models lack long-term memory and consistency [9][10] - The proliferation of homogeneous AI characters has diluted user interest, making it difficult for these products to establish emotional connections [11] - Successful AI characters, like "Ani" from Grok, demonstrate the importance of unique and engaging characters in driving user engagement [12][14] Financial Viability - The AI social sector is witnessing a decline in investment interest, with no significant funding cases emerging in 2023 [15][16] - The financial performance of leading AI social products is underwhelming, with Character.AI's 233 million monthly active users generating only $1.67 million in annual revenue, indicating a low user monetization rate [16] - High operational costs and low user retention rates are leading to unsustainable business models for many AI social applications [16][17] Future Outlook - The AI social market may face extinction before the next technological breakthrough, as current products struggle to find a viable path to profitability [17] - Despite the current downturn, the inherent demand for emotional companionship suggests that future technological advancements could reignite interest in AI social applications [19]

Core Viewpoint - The AI social application sector is facing a significant survival crisis in 2025 after a period of rapid growth and intense competition in 2024, with major players experiencing drastic declines in user engagement and revenue challenges [3][7]. Group 1: Market Dynamics - In 2025, leading AI social applications in China, such as Byte's Cat Box and MiniMax's Starry Sky, saw their daily downloads plummet from over 20,000 to below 7,000, representing a decline of more than 65% [3]. - Character.AI, a prominent overseas AI social application, boasts 200 million monthly active users but struggles with monetization, generating only $16.7 million in annual revenue [5]. Group 2: Challenges in AI Social Applications - The AI social application market faces several challenges, including low technical barriers leading to intense competition, where small teams and individual developers can quickly launch new applications [8]. - The threat of substitutes is high, as large models like ChatGPT already offer chat functionalities, reducing the need for users to download additional AI social applications [9]. - The industry suffers from severe homogenization, with many applications lacking unique features and relying on price wars, leading to a detrimental cycle of low-quality competition [9]. - High computational costs and low user willingness to pay further complicate the business model, as users have become accustomed to free services [9]. Group 3: Future Directions for AI Social Applications - AI social applications must decide whether to continue focusing on traditional virtual character chat apps or to pivot towards content transformation and service expansion to find new value [14]. - Potential strategies include shifting from "one-way emotional companionship" to "content consumption and co-creation," allowing users to engage more deeply with virtual characters [14]. - Transitioning from a consumer-focused (ToC) model to a business-focused (ToB) model could enhance profitability by addressing clear pain points for enterprise clients [14]. - Emphasizing high-quality content and unique character designs rather than expanding a generic character pool could lead to more successful applications [15]. - The ongoing demand for social interaction suggests that there remains a viable market for innovative AI social solutions, as evidenced by new entrants like Elon Musk's xAI [15][17].

Core Viewpoint - The rapid development of AI technology has led to the emergence of AI companions, enhancing human-machine interaction and providing emotional value to users [1][10]. Industry Overview - The AI emotional companionship market in China is projected to reach 1.211 billion yuan in 2024, potentially exploding to 3.866 billion yuan in 2025, and is expected to exceed 59.5 billion yuan by 2028, with a compound annual growth rate of 148.74% [7]. - Companies like Minimax, ByteDance, and Tencent are also developing similar AI companionship products, indicating a growing interest in this sector [7]. Product Development - The product "Doudou Game Partner" by Beijing Xinying Suixing Technology Co., Ltd. allows users to interact with AI characters in gaming scenarios, aiming to establish emotional connections through gameplay [2][4]. - The company focuses on creating high-quality AI characters with auditory and visual feedback to enhance user experience [8]. Ethical Considerations - The rise of AI companions has sparked discussions about ethical boundaries, particularly concerning their impact on youth and real-life relationships [1][10]. - Experts emphasize the need for regulatory frameworks to ensure that AI technologies promote social interaction rather than deepen isolation [10][11]. Future Outlook - The potential for AI to meet human emotional needs is seen as a natural evolution, with expectations that future AI collaborations will expand emotional connections [11].

Core Insights - The explosive growth of AI social applications, particularly those focused on emotional companionship, has faced a significant survival crisis in 2025 after a competitive boom in 2024 [1][3] - Major AI social applications in China, such as Byte's Cat Box and MiniMax's Starry Sky, have seen daily downloads plummet from over 20,000 to below 7,000, a decline of more than 65% [1] - Investors are pulling back, and large companies are adopting a wait-and-see approach due to challenges in monetization and high operational costs [3] Industry Challenges - The AI social application sector faces several evident challenges, including low technical barriers leading to intense competition, where small teams and individual developers can quickly launch new applications [4] - High threats from substitutes like ChatGPT, which already provide chat functionalities without requiring additional downloads, make it difficult for dedicated AI social apps to attract users [4] - The industry suffers from severe homogenization, with many applications competing on price and lacking unique features, leading to a vicious cycle of low-quality offerings [4][6] Market Dynamics - Character.AI, a leading overseas AI social application, boasts 200 million monthly active users but struggles with monetization, generating only $16.7 million in annual revenue [5] - The high cost of computational resources and low willingness to pay among users further complicate the financial viability of AI social applications [6] Strategic Directions - AI social applications must consider pivoting from traditional "virtual character companionship" to more diverse content consumption and co-creation experiences to meet user needs [9] - Transitioning from a consumer-focused (ToC) model to a business-focused (ToB) model could provide better ROI by addressing clear pain points for enterprise users [9] - Emphasizing high-quality content and unique character designs rather than expanding a generic character pool may lead to more successful products [10] Future Opportunities - The AI social sector is still in its early stages of interactive experience development, and future success will depend on addressing diverse emotional needs of users [12] - New entrants continue to emerge in the market, indicating ongoing interest and potential for innovation despite current challenges [10][12]

Core Viewpoint - The AI social application sector, which experienced explosive growth in 2023, faces a severe survival crisis in 2025 due to intense competition and challenges in monetization [1][2][6]. Group 1: Market Dynamics - In 2024, leading AI social applications in China, such as Byte's Cat Box and MiniMax's Starry Sky, saw their daily downloads plummet from over 20,000 to below 7,000, a decline of more than 65% [2]. - Character.AI, a leading overseas AI social application, boasts 200 million monthly active users but generates only $16.7 million in annual revenue, indicating significant challenges in commercialization [4]. Group 2: Challenges Faced by AI Social Applications - The low technical barrier and lack of competitive moat allow personal developers and small teams to quickly launch new applications, leading to fierce competition [8]. - The threat of substitutes is high, as large models like ChatGPT already provide chat functionalities, reducing the need for users to download additional AI social applications [9]. - The industry suffers from severe homogenization, with many applications competing on price and lacking unique user experiences, leading to a vicious cycle of low-end competition [9]. - High computational costs and low user willingness to pay hinder profitability, as evidenced by Character.AI's struggle despite its large user base [10]. Group 3: Potential Paths for AI Social Applications - AI social applications must decide whether to continue focusing on virtual character chat apps or to pivot towards content transformation and service expansion to find new value [14]. - Transitioning from a consumer-focused (ToC) model to a business-focused (ToB) model could provide better ROI by addressing enterprise needs such as language practice and employee mental health support [15]. - Emphasizing refined content and creating high-quality, engaging narratives can enhance user retention, as seen in successful otome games [16]. - The future of AI social applications lies in their ability to address diverse emotional needs of individuals, suggesting a potential for niche and vertical market opportunities [18].

Core Viewpoint - The article discusses the emergence of AI companionship as a controversial yet rapidly growing sector, particularly highlighted by Elon Musk's xAI and its chatbot Grok, which has introduced a "companions" feature based on the Grok4 model [1][2]. Group 1: AI Companionship Market Dynamics - The AI companionship market is gaining attention, with Musk's project aiming to compete against OpenAI, indicating a significant shift towards emotional engagement in technology [2]. - The launch of the gothic character Ani has quickly captured social media interest, demonstrating the potential for AI companions to fulfill emotional needs and create user engagement [4]. - The contrasting approaches of different projects, such as Musk's Grok and Mihayou's game "Whispers From The Star," highlight the diverse user demands within the AI companionship space [6]. Group 2: Software Innovations and User Engagement - The success of Character.AI in 2022 showcased a previously overlooked market where users are willing to pay for virtual emotional connections, combining large model technology with role-playing [9]. - Replika, established in 2016, emphasizes the identity of an "AI friend" rather than just role-playing, adapting to user interactions to create a personalized experience [10]. - Character.AI is projected to have over 28 million monthly active users by 2025, with revenue expected to rise from $15.2 million in 2023 to $32.2 million in 2024, reflecting a growth rate exceeding 100% [13]. Group 3: Hardware Developments and Challenges - As software competition becomes saturated, hardware innovations like AI companion toys are emerging as new avenues for growth, with products like "Ah Beibei" and "Loona" designed to provide emotional support and interaction [16][17]. - The Japanese brand LOVOT focuses on creating emotional attachment through non-verbal interactions, achieving significant sales despite a high price point [19]. - The entry of major players like Musk into the AI companionship market raises questions about the sustainability and depth of emotional engagement that technology can provide [20]. Group 4: Regulatory and Ethical Considerations - Content regulation remains a critical issue, with concerns about the effectiveness of filtering mechanisms in AI companions like Grok, particularly regarding sensitive content [20]. - The potential for user data from intimate conversations to be included in training datasets raises privacy and compliance issues, especially in light of EU regulations [20]. - The current limitations in AI's emotional understanding highlight the need for technological advancements and a balance between innovation and regulation for the market to mature [21].