AI Social Networking: Assessing Global Business Potential

Recently, several overseas AI social applications such as LoveyDovey and SocialAI have gained popularity, prompting many netizens to follow suit and interact with virtual characters. Previously, numerous Chinese manufacturers have launched AI social products overseas to test the waters, including MiniMax's Talkie, Kunlun Wanwei's Linky, and Zuoyebang's Poly.AI, with some applications leading globally. Domestic AI social products are also emerging continuously, with Tencent, ByteDance, Soul, Douyin, Xiaohongshu, Meituan, and Yuewen all placing bets.

"The main differences between AI social products and traditional social products lie in intelligent interaction, personalized experience, and interaction with virtual characters," said Wang Zhenlong, founder and CEO of Yuntai Capital, in his analysis of new media. The strong natural language understanding capabilities of large models, the unpredictability of generative AI responses, combined with users' needs for chat companionship, make AI social an important application scenario for AI at present. Its market potential at home and abroad is becoming apparent, especially in the overseas market, where some domestic AI social products have shown strong competitiveness.

Advertisement

With the advancement of AI technology, innovative AI social play methods continue to emerge, and business models are also being continuously explored. In the future, whether the status of social giants can be shaken, and who is building a moat?

1. AI social is on fire, and domestic applications are accelerating to catch up

In August this year, Character.AI, an artificial intelligence chatbot startup "sold" to Google, launched customizable AI characters, providing emotional value through companionship chat, with both social and gaming attributes. According to a report from venture capital firm a16z, as of January 2024, among the top 50 independent monthly visits of generative AI products, Character.AI ranked third in the web version, just behind the general chatbot ChatGPT and Gemini.

However, AI social applications show better user stickiness. For example, according to SENSOR TOWER data, in March 2024, Character.AI's DAU/MAU reached about 31%, while ChatGPT's figure was only 15%. Although Character.AI has fewer users than ChatGPT, indicators such as user retention rate, DAU/MAU, user usage time, and usage frequency are all much better than the latter.

New media has noticed that domestic AI social newcomers are not lagging behind. In the list of Chinese AI application products going overseas in August 2024,统计 by Feifan Chanyan, several AI companion social products such as Talkie, Poly.AI, Linky, and AnyDoor have squeezed into the top 20 downloads.

Looking at the companies behind them, most are not traditional social giants. Talkie, produced by Minimax, focuses on AI chat + card drawing mode, with a download volume of 2.35 million in August, a month-on-month increase of 31.63%, ranking sixth among Chinese AI APP products going overseas. Poly.AI comes from the education and training giant Zuoyebang, initially positioned as a language learning application, and later transformed into an AI companion, with a download volume of 2.2 million in August, a month-on-month increase of 38.76%, ranking tenth. According to the TOP10 AI product downloads in the United States released by Sensor Tower in the first half of 2024, Talkie is 4th, and Poly.AI is 9th.

ByteDance is one of the few social giants that have squeezed into the forefront of AI social products. Its product AnyDoor (formerly BagelBell) is positioned as an AI plot interaction platform, supporting user interaction with AI characters, with a download volume of 570,000 in August, a month-on-month increase of 6.38%. In comparison, Linky, a subsidiary of Kunlun Wanwei, has a similar business direction, also enabling users to interact more deeply with AINPC through AIGC technology, integrating story, script creation, and chat interaction functions in the App, with a download volume of 780,000 in August, and daily active users breaking through 500,000.

Compared to the aforementioned products, MyShell and joyland mainly focus on overseas chat application services, respectively from MyShell and Xihu Xinchen, two startups, with download volumes of less than 100,000 in August, but month-on-month growth reaching 322.97% and 564.29%, respectively.According to the analysis by Yang Xiaofeng, an analyst at Huafu Securities, based on the 7mai rankings, Talkie has consistently ranked higher than Character.AI in the US entertainment app category since September 2023. At the same time, the app's daily active users and average daily usage time are closely following Character.AI, even surpassing it at times. In his view, compared to Character.AI, Talkie has improved user retention and chat duration by capturing user interests.

Dongwu Securities also stated in its research report that Character.AI satisfies a relatively broad range of user needs, with anime games and assistants being the most popular; whereas Talkie and Linky cater more to companionship and romantic needs, with a significantly higher proportion of male users than Character.AI.

In terms of user gameplay, compared to Character.AI's simple character list mode, Talkie has adopted a left-right swipe mode with character visualization, offering a stronger sense of immersion; Linky's mode is similar to TALKIE, with its highlight being a personalized recommendation mode. In terms of creation mechanisms, Character.AI has a strong ability to define characters, Talkie has a refined character setup, and AI-generated balance threshold and effect issues; Linky has a relatively low creation threshold, and creator incentives are expected to drive an increase in supply. In terms of monetization mechanisms, Character.AI, Talkie, and Linky all have subscription plans. Talkie and Linky also have a card-drawing mechanism, which, benefiting from a stronger sense of user immersion, results in significantly higher MAU ARPU than the former.

II. Personalization + Diversification, Continuous Emergence of Innovative Gameplay

When discussing the trends and changes in Chinese social products going global, Wang Zhenlong believes that the market size growth, the increase in audio and video social interactions, and the increasingly distinct characteristics of different markets are becoming more evident. For example, the Middle East market is keen on voice social products, while the European and American markets prefer short video and other social products. Compared to other AI field export models, AI social export focuses more on technological innovation and user experience, pays more attention to cultivating user stickiness, and considers the cultural differences and user habits of the target market more.

Further sorting out domestic and overseas AI social products such as ByteDance's Cat Box, Yuewen's Dream Island, and MiniMax's Star Field reveals that with the advancement of AI technology, innovative gameplay is continuously emerging. Taking Linky as an example, its character and story models assemble outlines and stories of various lengths with tags, engaging in long-lasting role-playing interactions with users. Currently, the product integrates story, script creation, and chat interaction functions within the App, supporting users to generate Chat scripts with one click.

Kunlun Wanwei told New Media that Linky has built a UGC character ecosystem closed loop, with an average of tens of thousands of UGC contents per day, covering a rich range of plot NPCs and character settings. This ecosystem provides users with a variety of role-playing options and motivates users to continuously improve content quality through a commercialized revenue-sharing system, ensuring the platform's vitality and sustainable development.

There are also companies that run multiple interaction innovations in parallel. In terms of Soul's exploration of AIGC's implementation in social scenarios, there are mainly two main lines: one is to assist in social interaction, that is, to let AI intervene in the discovery, ice-breaking, and sedimentation of relationships, to improve the quality and efficiency of relationship establishment, and the new functions of in-station AI recommended relationships and AI-assisted conversations are attempts in this direction. The second is to build a new scenario for human-computer interaction, allowing users to interact with AI in a more warm, immersive, and interesting way, and the anthropomorphic dialogue robot "AI Gou Dan" is one of the applications.

"The rapid development of natural language processing and generative AI has provided new possibilities for AI social products," Wang Zhenlong believes that with the increasing demand for personalized and intelligent social experiences from users, the AI social market will usher in greater growth space. AI social products provide enterprises with the opportunity to explore new business models and can create new sources of revenue. However, how AI applications make money is a global problem, and domestic developers need to find more effective monetization paths.

III. Solving the Money-Making Problem, How Much Room for Business Model Innovation?Currently, the main monetization methods for AI social applications include paid subscriptions, card drawing, and advertising. Many enterprises have expressed to new media that AI will further open up commercial space, which not only means the opening of traditional commercial spaces such as value-added services, advertising, and e-commerce but also the emergence of new business models.

Qin Heping, an analyst at Guotai Junan Securities, believes that in the short term, AI helps to improve the quality and efficiency of advertising, making this business model evergreen. In the long term, AI helps to solve the matching efficiency problem in social networks, which is expected to lead to a variety of business models. For example, AI avatars from both parties can first conduct preliminary communication to achieve rapid mutual understanding and break the ice quickly, which is expected to lead to business models such as marriage matching, monetizing through one-time matching fees or membership subscriptions. It can also be applied to the broad social network based on people to realize value interaction. For instance, AI can achieve precise matching between users and restaurants based on users' dietary restrictions, allergies, and taste preferences, monetizing through charging merchants a commission or charging users a one-time matching fee.

Soul also revealed to new media that at this stage, Soul is trying to provide brand digital solutions for the B-side. For example, brands also need to connect with people. When platform users communicate with NPCs and mention wanting to drink milk tea, it can link to milk tea services. If AI technology is used, the production efficiency changes, and it is no longer limited to a single type of output, including but not limited to virtual avatars, music, and emoticons. More people can participate and bring more possibilities for creative personalized services, thus becoming producers and receiving a share.

"As an investment institution, the key elements of the business model we value most include technological innovation ability, user stickiness, monetization ability, market positioning, and user demand satisfaction. These elements of the business model are interrelated." In Wang Zhenlong's view, the subsequent upgrade of multimodal capabilities of large models and more innovative gameplay is expected to drive the continued growth of AI social users. Drawing on the experience of the development of mobile Internet, stronger user stickiness and usage time are expected to give the platform stronger user monetization capabilities. At the same time, it is also necessary to pay attention to relevant laws and regulations, data security, and privacy protection.

Regarding competition, Soul believes that from the perspective of the development of AI industries at home and abroad, China may still need to catch up in terms of technical capabilities, but domestic enterprises have advantages in scenarios and data. As an explorer at the application level, Soul itself naturally has the advantage of application scenarios, avoiding the current industry's common embarrassing problem of "looking for nails with a hammer."

"AI continues to evolve, and in the future, the truly valuable ones are the business models that provide irreplaceable application ecosystems in specific fields." Guotai Junan Securities analyzed in its research report that vertical platforms represented by Character.AI first make products, and social giants first make models and ecosystems. Social giants have user base + model capabilities + application ecosystems, and their advantages can be transferred to vertical scenarios, making it natural to make AI virtual social products. For example, Tencent has a rich reserve of IPs, and the character IPs in game products can be applied to AI virtual social products.

In their view, the commercialization at this stage is still in its early stages, and the advertising revenue model has not yet been proven. Character.AI not only faces the high cost of training and operating large models but also faces competition from social giants entering the chatbot track. Character.AI's first-mover advantage is gradually weakening, and social giants are catching up, and they will still dominate the AI social track in the future.

Wang Zhenlong believes that the key factors for the success of an AI social product include technological innovation and user experience, strong user stickiness, innovation and sustainability of business models, market positioning and demand satisfaction, and data security and privacy protection. Domestic Internet giants usually have rich user data and strong algorithm R&D capabilities, which enables them to use AI technology to optimize social experiences and take the lead. However, small and startup companies also have broad opportunities in the AI social product industry, especially in technological innovation, differentiated positioning, and business model innovation.