Original · Unique Research · 2026-02-04
Historical edition: This is the full textual English rendition of Unique Research’s February 4, 2026 article discussing December 2025. Product capabilities, rankings, growth explanations, forecasts, institutional descriptions, methodology and accuracy claims remain the original author’s statements at that time, not independently verified current findings. The source’s unique-visitor and subscription-revenue definitions, including its ARR label, are reproduced as its own methodology.
Reading note: The source calls active users divided by visits a “conversion rate” and interprets it as retention; that ratio alone does not establish cohort retention or paid conversion. Likewise, a 335.36% increase in downloads does not establish an equivalent increase in distinct users. Claims about every edit supplying model-training data, causes of churn, network effects, market dominance and comparative lifetime value are the author’s interpretations, not independently verified operational facts. A negative category average does not by itself establish that most products declined; rankings and percentages should be read within their stated Web/app and metric scopes. The 51.2% share concerns emotional-companionship traffic in this sample, not the entire AI market; the 234.1% app growth and -0.58% Web-context figure have different scopes. In particular, the active-user/visit ratio cannot establish that 64 out of 100 visitors subsequently stayed.
When Virtual Lovers Dominate Half the Market,
What Is Chinese AI Going Overseas Actually Selling?
An In-Depth Look at December 2025 Data on Chinese AI Products Going Overseas
Unique Research | AI Top 100, with 8 Rankings at the End
01 / Why Did an Education Company Rank First in Overseas Traffic for Chinese AI Products?
In December 2025, PolyBuzz topped the WEB traffic ranking for Chinese AI products going overseas with 42.26 million monthly visits, up 25.89% month over month. What does that number mean? It exceeded Alibaba's Qwen, the combined total of every AI product from ByteDance, and SeaArt AI from Haiyi Interactive Entertainment (海艺互娱). A company that began in K12 education used a virtual-companion product to secure the most traffic in the fiercely competitive market for Chinese AI going overseas.
42.26 Million Visits
PolyBuzz Ranks First in Monthly Visits
This is not an isolated case. Across the emotional-companionship field, the top three products held 85.8% market share. PolyBuzz alone took 51.2%. This near-monopolistic concentration appears in no other AI field. Image generation has 11 products dividing the traffic, video generation has 16 competitors, and even the agent category has 7 players. Emotional companionship alone presents a winner-takes-all picture.
The question is: why emotional companionship? Why not image generation, which appears more essential; video generation, which is more spectacular; or agents, which appear more imaginative? While everyone discusses AIGC's productivity value, the products actually making money and gaining users overseas are those selling virtual relationships.
02 / The Truth About Traffic: Who Is Leading Chinese AI Overseas?
On the WEB side, Chinese AI products going overseas recorded 450 million total visits, across 100 products in December 2025. Yet those 450 million visits were distributed extremely unevenly. Although emotional-companionship products numbered only 7, they took 82.53 million visits, representing 18.4% of the total. By comparison, 11 image-generation products recorded 71.35 million visits, while 16 video-generation products recorded 58.67 million visits. On average, traffic per emotional-companionship product was 1.8 times that of image-generation products and 3.2 times that of video-generation products.
450 Million Visits
Total Visits to Chinese AI Products Going Overseas
The gap was more obvious on the APP side. ByteDance's Hypic recorded 18.31 million monthly downloads, up 164.85% month over month. It is an image-editing tool, but its growth was far slower than Kuaishou's KLING AI, which became the fastest-growing APP product with a 335.36% month-over-month increase. Average growth in the video-generation category reached 234.1% on APP, more than 6 times that of image editing.
On the surface, image-editing tools led in downloads, but video generation had already overtaken their momentum. Users were voting with their feet as their interest shifted from static images to dynamic video. This was not merely the result of technological progress, but a signal that demand was shifting from "beautification" to "expression." A retouched photograph can be posted on social media, but a creative short video can spread farther.
03 / Emotional Companionship's Monopoly Logic: Why Is There Only One Winner?
First, the data. PolyBuzz's 42.26 million visits and 51.2% market share were not bought by burning money. Its month-over-month growth of 25.89% was not the fastest in emotional companionship—HiWaifu grew as much as 38.18%, while Talkie grew 24.15%. But PolyBuzz started from a large base, and its incremental growth equaled the combined growth of every competitor.
Where did this advantage come from? Not technology or brand, but network effects. The essence of an emotional-companionship product is a relationship, and once formed, a relationship is difficult to switch. Users invested time, emotion, and conversation history in PolyBuzz, none of which could be transferred to another platform. In other words, users were not using a product, but maintaining a relationship. That stickiness is stronger than for any tool product.
CrushOn.AI ranked second with 22.64 million visits and a 27.4% market share. Its gap from PolyBuzz looked large, but its growth was not weak, rising 14.02% month over month. The question was whether that growth could continue. CR3, the concentration of the top three emotional-companionship products, had reached 85.8%, leaving only 14.2% of the market. Emochi in fourth place and Talkie in fifth must either accept being small players or find differentiated breakthroughs.
Where is the differentiation? MiniMax's Talkie offered one answer: although it had only 4.59 million WEB visits, its APP growth reached 124.68%. This showed that it had found a mobile position. Users engage with virtual companions more frequently and in more scenarios on phones, from fragmented chats to companionship before sleep, making mobile naturally more suitable for such products. Talkie captured this opportunity. Although PolyBuzz dominated WEB, its APP ranking was only ninth, with 2.27 million downloads, far below the leaders.
04 / ByteDance's Two Paths: Tools or Ecosystem?
ByteDance is the most active player in Chinese AI going overseas, without exception.
On WEB, 4 products appeared: Gauth, education, ranked 9; Dreamina, image generation, ranked 12; Dola AI, personal assistant, ranked 43; and TRAE, coding assistant, ranked 48. On APP, there were 3: Hypic, image editing, ranked 1; Dola AI, chatbot, ranked 2; and Gauth, education, ranked 13. The product matrix shows ByteDance doing one thing: entering through tools and feeding data into models.
18.31 Million Downloads
Hypic Monthly Downloads
Hypic is a typical case. It recorded 18.31 million monthly downloads, up 164.85% month over month, and 48.67 million monthly active users. What does that mean? Its active users alone equaled the total for every product ranked from 3 through 10. More importantly, Hypic is not merely an image-editing tool; behind it is ByteDance's Doubao foundation model. Every image users generate and every edit they make provides training data for the model.
But this path has a problem: tools have a low ceiling. Dola AI had 55.93 million monthly active users, ranking first on APP, but its downloads fell 22.7% month over month. Users' novelty toward chatbots is fading. They need not more functions, but clearer use cases. Dola AI attempted to become an all-purpose assistant, with the result that it could do a little of everything but nothing deeply enough.
ByteDance's other path is more hidden, but perhaps more important: Gauth. This AI tutoring tool for students recorded 12.43 million WEB visits and 26 million APP monthly active users. Its data were not top-tier, but its users were highly valuable. AI conversations in education require higher quality and tolerate fewer errors, so the data users produce here are far more valuable for model training than ordinary chat. ByteDance is using Gauth to build a data flywheel in the education vertical; once proven, it can be replicated across more verticals.
The problem is that Gauth's growth stalled. WEB declined 19.42% month over month and APP fell 76.85%. Users were leaving for a simple reason: too many free users and overly difficult paid conversion. Education tools address clear pain points, but users have little willingness to pay, especially overseas. ByteDance must balance product value and commercial model, or this data flywheel will not turn.
05 / Kuaishou's KLING AI Explosion: Has Video Generation Reached an Inflection Point?
If ByteDance is laying out a product matrix, Kuaishou is betting on a single-point breakthrough.
KLING AI recorded 13.46 million WEB visits, ranked 7, and grew 33.79% month over month. On APP, it had 2.93 million downloads, up 335.36% month over month, making it the fastest-growing product in the entire December. What did that increase mean? The source interprets this as its user base reaching more than 4 times its previous size in one month.
335.36%
KLING AI APP-Side Month-Over-Month Growth
KLING AI was not the first video-generation product, but it may have been the first to raise the experience above the passing line. Users care most about three things: generation speed, visual quality, and operational barriers. KLING AI reached the industry's front ranks on all three. Generation was fast enough for real-time previews, image quality was commercially usable, and the interface was simple enough for beginners. Together, these three factors formed the foundation for explosive growth.
More important was timing. In December 2025, average growth in video generation was -0.58%, meaning most products declined. KLING AI grew against the trend not because it was vastly stronger than others, but because it found the right rhythm. Demand for video generation always existed, but prior products were too expensive, too slow, or ineffective. KLING AI resolved all these problems at once and naturally became users' first choice.
This sounded an alarm for other players. PixVerse ranked third, with 7.08 million downloads and growth of 133.83%, also an impressive performance. But its WEB growth was only 9.08%, indicating that its users mainly came from mobile. The video-generation battlefield is shifting from desktop to mobile; whoever establishes an advantage on mobile can capture the next growth dividend.
06 / Dark Horses and Reshuffling: Where Is the Formula for Growth?
Rankings always contain unexpected names.
Flowith, classified here as an intelligent-search tool, recorded 1.61 million WEB visits and grew 157.57% month over month. It had the highest growth rate among all products with more than 1 million visits. What is intelligent search? Put simply, it redefines the search experience with AI. Users do not enter keywords; they ask questions directly, and AI understands intent, integrates information, and provides an answer. This may sound no different from traditional search, but the experience is completely different. Traditional search gives you links to inspect yourself; intelligent search gives you the answer directly and allows follow-up questions.
Flowith's growth illustrates a trend: users increasingly demand efficiency in information acquisition. They do not want to waste time filtering, comparing, and judging. They want fast, accurate, trustworthy answers. Intelligent search grew by an average of 47.99%, the fastest of all categories. This was not a coincidence, but a concentrated explosion of user demand.
Oreate AI was another dark horse, a productivity tool with 1.74 million WEB visits and month-over-month growth of 117.14%. Its positioning was clear: help users manage tasks, organize information, and optimize workflows. That sounds conventional, but it grew faster than every comparable product because it addressed a pain point: information overload. Modern people process 10 times as much information as 10 years ago, yet information-processing tools remain where they were 10 years ago. Oreate AI uses AI to filter useless information, extract key content, and generate summaries automatically, precisely what users need most.
07 / The Secret of Conversion: Traffic Does Not Equal Retention
On WEB, conversion from visits to active users varied enormously. Fotor had 13.61 million visits and 8.67 million active users, for a conversion rate of 63.7%. What level is that? Among every 100 visitors, 64 remain and continue using it. By comparison, PolyBuzz had 42.26 million visits and 3.32 million active users, for a conversion rate of only 7.86%.
71.58%
123RF Conversion Rate
We all know traffic quality matters more than quantity. Fotor is an image-editing tool. Users arrive specifically to edit pictures, so the purpose is clear and the path is direct. Visitors may come to PolyBuzz merely from curiosity, with only a minority truly willing to form a relationship. But once that minority remains, their lifetime value far exceeds that of tool users.
123RF had one of the highest conversion rates, with 8.55 million visits, 6.12 million active users and a 71.58% conversion rate. It is an image-asset platform; users come to download assets, so demand is rigid and the scenario fixed. Such products naturally have high conversion, but an obvious ceiling: users do not come every day and open them only when needed.
By comparison, KLING AI's conversion rate of 35.22% looked low, but momentum was strong. Trying video generation carries greater costs; users need time to learn, experiment, and adjust before producing a satisfying work. Most people give up during the process, and only a minority persist. But once that minority masters the techniques, both usage frequency and willingness to pay rise significantly.
08 / Category Temperature: Who Is Heating Up, and Who Is Cooling Down?
On WEB, intelligent search grew by an average of 47.99% month over month, 3D modeling by 20.49%, and emotional companionship by 17.97%. These three categories performed strongly in December, reflecting structural changes in user demand. Intelligent search represented the pursuit of information efficiency, 3D modeling the need for creative tools, and emotional companionship acceptance of virtual relationships.
By comparison, education declined by an average of -17.00%, mind mapping by -12.76%, and agents by -8.20%. These fields cooled not because demand disappeared, but because products were highly homogeneous and user choices converged. Education tools all offered AI tutoring, mind maps all offered automatic generation, and agents all offered multiround conversations. Without fundamental differences in function, users naturally concentrated around the leaders.
The trend was more extreme on APP. Video generation grew by an average of 234.1%, the only category to break through 200%. Emotional companionship grew 28.19%, and image editing 34.69%, both remaining positive. But education declined by an average of -52.01%, almost halving. This was not a problem with individual products, but the entire category. Mobile users had little willingness to pay for education tools, commercialization was difficult, and developers began leaving the market.
Video editing grew by an average of 13.00%, which looked good, but internal divergence was severe. YouCut grew 24.56% and VivaCut 22.69%, while other products declined. The barrier to video editing is falling. Users do not need professional tools; they need simple, fast, effective solutions. Whoever delivers all three can capture the growth dividend.
09 / What Is Being Redefined Is Not Technology, but Relationships
Returning to the original question: why emotional companionship?
Because AI is redefining human-machine relationships. In the past, we treated computers and phones as tools and put them down after use. Now, we are beginning to treat AI as a partner for conversation, confidences, and advice. This is not a byproduct of technological progress, but an inevitable result once technology reaches a certain stage. When AI is intelligent, natural, and understanding enough, it ceases to be merely a tool and becomes a new form of relationship.
PolyBuzz's success lies not in advanced technology, but in capturing the trend's early opportunity. Users are there not to complete tasks but to gain companionship. This demand was magnified after the pandemic, reinforced after fatigue with social media, and redefined as real-world relationships became more costly. Virtual companions do not replace real relationships, but supplement them. They provide a low-cost, low-risk, highly controllable form of emotional connection.
ByteDance enters through tools, while Kuaishou breaks through with video generation. Both are searching for paths to AI commercialization. Yet the end point of commercialization is not accumulating functions, but establishing user relationships. How many images Hypic generates is unimportant; what matters is whether users open it every day. How many special effects KLING AI can produce is unimportant; what matters is whether users treat it as a creative partner.
From this perspective, competition among Chinese AI products going overseas has only begun. The current ranking reflects allocation of the first wave of dividends: whoever built first, launched first, and captured traffic first gained an advantageous position. But the second wave will be more brutal because user choices will converge, category concentration will rise, and opportunities for latecomers will diminish.
Intelligent search is heating up, video generation is exploding, and emotional companionship is consolidating. These three categories represent three models for Chinese AI going overseas: productivity tools, creative platforms, and relationship products. Which will become mainstream? The answer may lie not in technology, but in users' choices. When users turn AI from a tool into a partner, and when AI moves from an assistant to the star, the entire industry's narrative will be rewritten.
The data for December 2025 are a snapshot, reflecting not only the present but signals for the future. Can ByteDance's tool matrix become an ecosystem advantage? Can Kuaishou's KLING AI move from a single-point breakthrough to success across the field? The answers will gradually become clear over the next several months.
One point is certain: what Chinese AI going overseas sells is not technology, but relationships. Whoever becomes indispensable to users can win in the end.
Unique Research—An Authoritative Institution for Global AI Market Intelligence
Details of the December 2025 AI Top 100 published by Unique Research follow:
Outbound Chinese AI WEB Ranking: Top 100 outbound Chinese AI Web products by visits and monthly active users.
Outbound Chinese AI APP Ranking: TOP30 outbound Chinese AI App products by downloads and monthly active users.
Our Influence and Recognition
Unique Research's data and insights have earned broad trust and citation from leading investment institutions, academic organizations, and media in China and abroad, becoming an important benchmark for market analysis, investment decisions, and trend assessments. Our partners and citing organizations include, but are not limited to:
Securities firms and investment banks: Sinolink Securities, China Securities (中信建投), Huayuan Securities, Haitong Securities, and others.
Authoritative media: China Securities Journal, TMTPost, Jiemian News, and others.
We firmly believe that only data capable of withstanding validation by core industry participants possesses long-term value.
The Power of Open Source and Openness
We believe in the power of open source and openness. Unique Research hereby makes the following three commitments:
1. The AI Top 100 ranking will remain permanently open source and freely available, with the complete methodology and raw data supporting full reproduction.
2. Data sources for the AI Top 100 ranking are strictly limited to independent third-party monitoring platforms; companies' self-reported data are not accepted as evaluation evidence.
3. The AI Top 100 ranking system remains absolutely neutral and accepts no commercial partnership or payment that intervenes in ranking results.
How Should the AI Top 100 Ranking Be Used?
If you are an investor/analyst: this ranking provides a clean data benchmark that can withstand reverse-engineering verification, serving as a reliable foundation for building investment models, discovering below-the-surface projects, and conducting cross-validation.
If you are an AI entrepreneur/product manager: you can not only identify competitors precisely, but also reproduce our methodology in depth to analyze your own product's market performance and optimize its iteration.
If you are a researcher/student: its fully open-source nature makes this report an excellent free repository for academic research and market analysis. We encourage innovative research based on these data.
For the complete data behind the charts and rankings shown in the AI Top 100, visit the 100aiapps.cn website. All visible data can be copied and used free for noncommercial purposes such as personal learning and research. Commercial use requires prior consultation with Unique Research for authorization. We welcome ranked teams, subject to compliant use, to include their positions in business plans, recruitment materials, and other settings as authoritative evidence of their strength.
Join the AGI Think Tank Open-Source Community
The AI Top 100 is not merely a ranking, but an open-source community committed to promoting transparency in AI market data. We invite developers, data scientists, and industry analysts to:
1. Reproduce and validate: visit our AGI Think Tank to obtain the complete data and methodology, verify our results, and offer suggestions for improvement.
2. Build together: contribute code to help us expand data dimensions, such as API call-volume estimates and agent-monitoring models.
3. Exchange insights: join our AGI Think Tank discussion group and share distinctive insights derived from Unique Research data with peers.
One person's insight is limited; a group's wisdom is boundless. Together, let us define the standards for measuring the AI market.
I / Outbound Chinese AI WEB Ranking
1.1 Outbound Chinese AI WEB TOP 100 by Monthly Visits
Based on its independent research into the AI industry and continuous tracking of thousands of AI companies and nearly ten thousand AI products worldwide, Unique Research ranks the month's AI WEB products by cumulative visits and compiles the top 100 AI WEB products from Chinese or ethnic-Chinese entrepreneurs for overseas markets, as follows:
Visit the 100aiapps.cn website for the complete data; all visible data can be copied.
1.2 Outbound Chinese AI WEB TOP 100 by Monthly Active Users
Based on its independent research into the AI industry and continuous tracking of thousands of AI companies and nearly ten thousand AI products worldwide, Unique Research ranks the month's AI WEB products by active users and compiles the top 100 AI WEB products from Chinese or ethnic-Chinese entrepreneurs for overseas markets, as follows:
II / Outbound Chinese AI APP Ranking
2.1 Outbound Chinese AI APP TOP 30 By Monthly Downloads
Based on its independent research into the AI industry and continuous tracking of thousands of AI companies and nearly ten thousand AI products worldwide, Unique Research ranks the month's AI APP products by cumulative downloads and compiles the top 30 AI APP products from Chinese or ethnic-Chinese entrepreneurs for overseas markets, as follows:
2.2 Outbound Chinese AI APP TOP 30 By Monthly Active Users
Based on its independent research into the AI industry and continuous tracking of thousands of AI companies and nearly ten thousand AI products worldwide, Unique Research ranks the month's AI APP products by active users and compiles the top 30 AI APP products from Chinese or ethnic-Chinese entrepreneurs for overseas markets, as follows:
III / Research Statement
3.1 Data Notes
AI-product revenue data in this report, article, ranking, or chart are based on real-time tracking of traffic moving from official AI-product websites to payment gateways such as Stripe, combined with time-series analysis to examine historical visits, fluctuations in monthly active users and changes in pricing strategies, thereby constructing a dynamic revenue-estimation model. The model's core parameters are determined through a three-part validation system comprising paid-conversion-funnel monitoring, reverse calibration against corporate disclosures, and comparisons with industry benchmark conversion rates, and are promptly compared with and corrected against corporate financial reports, financing announcements, and authoritative media disclosures.
Sampling validation shows that the mean absolute error rate between predictions and actual data remains within ±10%, meeting industry-research accuracy standards.
AI-product traffic data in this report, article, ranking, or chart are based on multisource data integration and processing by Unique Research's intelligent algorithms. The monitoring scope is strictly limited to users' direct access through official applications and websites. Specifically, these data cover only user behavior involving direct visits to AI products through websites and native applications and exclude traffic generated through the following access methods: browser plug-ins/extensions, desktop client software, WeChat/Alipay mini-program ecosystems, embedded services on third-party platforms such as Discord, local deployment of open-source models, API calls, and other nondirect-access scenarios. The data focus on monitoring direct use of AI products through core user endpoints, aiming to reflect objectively mainstream users' direct terminal usage.
Given the complexity and dynamic nature of data collection and processing, as well as continuing changes in the market environment, the displayed data may contain some errors and omissions. These data should therefore be regarded as references for research and analysis, intended to provide users with a window into the AI-product market, rather than definitive evidence for making specific investment strategies or providing consulting advice.
Explanation of Data-Coverage Expansion:
We are actively developing and testing monitoring models in the following emerging fields and expect to incorporate them gradually into future quarterly reports:
AI agent ecosystems, or Agents
AI hardware-device sales and activity
Estimates of foundation-model API call volume
Please stay tuned.
3.2 Concept Definitions
Geographic Dimension
Overseas AI products: products founded by entrepreneurs or teams who are not ethnically Chinese and launched primarily for global markets, including but not limited to their home markets.
Domestic AI products: products founded by Chinese-national entrepreneurs or teams and launched primarily for China's domestic market.
Outbound Chinese AI products: products founded by ethnic-Chinese entrepreneurs or teams but principally targeting overseas markets outside China’s domestic market.
Functional Dimension
AI-native applications: applications that deeply integrate artificial-intelligence technologies and algorithms from the beginning of product design and whose core value, business processes, or user experience depends entirely on AI technology.
AI-feature applications: applications that integrate or embed AI technology into an existing business logic and application framework to enhance existing functions or provide entirely new ones.
AI-ecosystem applications: platform applications that support, connect, or facilitate communication among the AI ecosystem's technology researchers, application developers, service providers, end users and others.
3.3 Metric Notes
WEB Data Metrics
Visits: the total number of all visits to a website within a specified period, such as one month, used to measure traffic scale.
Unique Visitors: the number of distinct IP addresses visiting a target website within a specified period, such as one month.
Subscription revenue, or ARR: the total revenue obtained by a website during a specified period, such as one year, from subscription services provided to users.
APP Data Metrics
Downloads: the number of times users download an App within a specified period, such as one month.
Active Users: the number of distinct users who performed an activity within a specified period, such as one month.
In-app purchase revenue, or IAP: revenue generated when users purchase virtual goods, additional services, premium functions, and other items within an App.
3.4 Free-Use Statement
This report is published by Unique Research and its copyright belongs to Unique Research. Any Chinese-language reproduction or citation must identify the report source; overseas institutions seeking to reproduce or cite it should request authorization in advance. When citing data, use "Data source: Unique Research"; when directly citing charts, use "Data source: Unique Research; chart produced by Unique Research."
This report is Unique Research's independent original analysis as a third-party institution. Its content does not represent the position of any company and does not constitute investment advice for anyone. Investors must therefore note that Unique Research, its employees, and affiliated organizations bear no relation to any investment decisions made on this basis.
Where permitted by law, Unique Research and its affiliates may hold equity in companies mentioned in the report or provide, or seek to provide, fundraising, financial-advisory, or other related services to them. Employees may serve as directors of companies mentioned in the report.
08 / Appendix: Summary of Quarterly Trend Insights
Based on the current data, the Unique Research analyst team identified the following core trends:
The data on Chinese AI products going overseas in December 2025 provide a clear perspective: competition in AI is shifting from an expedition in computing power to a migration of desire.
When technical barriers are leveled, product moats no longer lie in algorithmic complexity, but in insight into human nature. Whether it is APP-side video generation's desire "to be seen," emotional companionship's dependence on "being present," or vertical tools' pursuit of "precise efficiency," all point to one overarching proposition: AI is no longer distant technology, but a nearby organ.
It is no longer a technology concept used for display, but an extension used to satisfy our everyday emotions, creative desires, and anxiety about efficiency.