★2025 Generative AI Consumer App Rankings: Ecosystem Stability and Global Competitive Landscape Analysis★
In the rapidly evolving landscape of generative AI technology, Andreessen Horowitz (a16z) has released its fifth edition of the “Global Top 100 Generative AI Consumer Apps Ranking,” providing a crucial window into industry development. This ranking incorporates 2.5 years of user behavior data, documenting the evolution of daily AI usage habits. As technology matures and markets consolidate, the generative AI application ecosystem is demonstrating new developmental trends.
Ranking Overview: Ecosystem Tendency Toward Stability
The most notable feature of this edition is the increasing stability of the overall generative AI application ecosystem. Compared to the fourth edition released in March 2025, fewer new applications have entered the rankings, reflecting the formation of a more settled market structure.
In the web application ranking, only 11 new entrants secured positions in the top 50 based on traffic growth, compared to 17 new entrants in the March 2025 ranking. This change indicates growing user loyalty among leading AI applications and an increase in market entry barriers.
The mobile application ranking saw 14 new entrants, significantly more than the web application category. This phenomenon primarily results from app stores strengthening their crackdown on “counterfeit ChatGPT” applications, creating market space for more original mobile applications and promoting healthier development of the mobile AI application ecosystem.
Notably, Google placed four products in the web ranking as independent entries for the first time. This development comes as Google further separated the domains of its AI products, allowing for independent traffic statistics. If these products had been tracked independently earlier, they would likely have appeared in previous rankings. Including these four Google products brings the total number of new entrants in this edition to 15.
Rising Stars: From Contenders to Mainstream
A16z published its second “Brink List,” featuring 10 companies that narrowly missed the top 100, with 5 web applications and 5 mobile applications. This list provides valuable insights into AI applications that may rise in the future.
From the previous web Brink List, Lovable successfully ascended to the main ranking, securing the 22nd position. This significant progress not only demonstrates Lovable’s development potential but also highlights the rising trend of AI-powered application generation platforms across the industry.
In the mobile application category, PolyBuzz and Pixverse from the previous Brink List also entered the core ranking, validating the Brink List’s predictive value for future trends.
Ranking Methodology Explanation
Understanding the ranking methodology helps interpret the significance behind the rankings. This ranking consists of two components:
- 🍄
Top 50 AI-native web products: Based on monthly unique visitor data from Similarweb - 🍄
Top 50 AI-native mobile applications: Based on monthly active user (MAUs) data from Sensor Tower
It’s important to note that products not originally AI-native but later adding significant generative AI features, such as Canva and Notion, were not included in this ranking. This selection criterion ensures the ranking more accurately reflects the development status of native AI applications.
Key Insight 1: Multiple Google Products Make Strong Showings
In this edition, Google placed four products in the web application ranking, marking the first time their traffic has been independently tracked and ranked separately. This development reflects Google’s comprehensive layout and rapid progress in the generative AI field.
Google’s general large language model (LLM) assistant Gemini ranked second in the web ranking, behind only ChatGPT, with approximately 12% of ChatGPT’s web traffic. This performance demonstrates Gemini’s strong competitiveness in the market.
Google’s other ranked products include:
- 🍄
AI Studio: Making its debut in the top 10. This developer-oriented platform provides a sandbox environment for developers to quickly start using Gemini models, including multimodal models (such as voice conversations and video streaming using Gemini Live).
- 🍄
NotebookLM: Ranking 13th. Originally part of Google Labs, it has now become an independent website. NotebookLM gained popularity a year ago and has maintained steady growth, with only a slight decline during summer months, possibly due to academic users’ seasonal vacations.
- 🍄
Google Labs: Ranking 39th. As Google’s consumer-facing AI experiment platform, Google Labs includes several innovative applications, such as Flow for experiencing the Veo 3 video model, virtual try-on application Doppl, AI coach application Portraits, and AI agent browser Project Mariner. After the release of Veo 3 in May 2025, Google Labs traffic surged by over 13%,创下 the largest monthly increase in a year.
On mobile devices, Gemini also performed well, ranking second behind ChatGPT but with a significantly smaller gap than on web platforms—its monthly active users nearly reached half of ChatGPT’s. Notably, Gemini usage is particularly high on Android devices, with nearly 90% of monthly active users coming from Android devices, compared to 60% for ChatGPT. This difference may relate to Google’s control over the Android ecosystem.
Key Insight 2: Intensifying Competition in General LLM Assistant Market
In the general large language model assistant market, while ChatGPT maintains its leading position, companies like Google, Grok, and Meta are narrowing the gap, creating a competitive landscape with multiple strong players.
X Corp.’s (formerly Twitter) assistant Grok performed prominently in this ranking, placing 4th in the web ranking and 23rd in the mobile ranking. Grok’s mobile growth is particularly notable, developing from a “cold start” state at the end of 2024 (when there was no official application) to now boasting over 20 million monthly active users, demonstrating impressive growth speed.
July 2025 marked a critical growth period for Grok, with mobile users surging nearly 40%. This growth primarily resulted from two important updates: the release of the new Grok 4 model on July 9, featuring enhanced reasoning capabilities, real-time search functionality, and tool integration; and the launch of AI companion avatars on July 14. Among these avatars, the anime character Ani, which includes NSFW (not safe for work) options, gained particular popularity initially.
In contrast, Meta’s AI assistant showed more moderate growth. Its general assistant Meta AI ranked 46th in the web ranking and failed to enter the mobile ranking. Meta AI first appeared at the end of May 2025 but grew much slower than Grok. This gap became more pronounced after June 2025, when users discovered privacy issues involving some posts appearing in public feeds, potentially affecting user trust in Meta AI.
Among other participants, DeepSeek and Claude have shown significantly slowed growth on mobile devices, with DeepSeek declining 22% from its peak. Meanwhile, Perplexity, along with Grok, continues to maintain strong growth momentum.
On web platforms, DeepSeek’s decline is more significant, dropping over 40% from its February 2025 peak, while Perplexity and Claude continue growing, demonstrating differences in user retention and market acceptance among products.
Key Insight 3: Global Influence of Chinese Domestic AI Applications
In this web ranking, three companies primarily serving Chinese users ranked among the top 20. These companies all have Chinese-language websites with over 75% of their traffic coming from China:
- 🍄
Quark: Ranking 9th, an “all-in-one” AI assistant launched by Alibaba, also ranking 47th in the mobile ranking - 🍄
Doubao: Ranking 12th, a general large language model product from ByteDance, performing better in the mobile ranking at 4th - 🍄
Kimi: Ranking 17th, a chatbot from startup Moonshot AI
These Chinese domestic AI applications’ presence in global rankings relates closely to China’s unique market environment. China has the world’s largest population, while many non-Chinese general LLM assistants like ChatGPT, Perplexity, and Claude are either blocked or restricted in China. Additionally, AI service providers operating in China must register and obtain licenses, requiring them to store data within China and comply with relevant review and content moderation regulations.
Despite these restrictions, some Chinese users access unregistered foreign products through VPNs or company/academic network gateways from within China, explaining why products like ChatGPT still have some traffic from China.
Beyond AI applications primarily serving Chinese domestic users, a significant proportion of products in the web ranking were developed in China but target global markets, with most users coming from other countries. Some tools are even blocked within China itself. Public data shows seven such companies, including Deepseek, Hailuo, Kling (video generation model), SeaArt (image generation), Cutout Pro (image editing), and Manus and Monica (productivity tools).
Chinese video generation models perform particularly well in global markets, often outperforming Western-developed models. This advantage stems mainly from two factors: China has more researchers focusing on video technology; and a relatively lenient intellectual property regulatory environment (possibly allowing training with copyrighted data) facilitates rapid technical iteration. Google’s Veo 3 is the first American model breaking this trend, partially trained using YouTube data.
This trend is even more evident in the mobile ranking. Among the 50 ranked applications, approximately 22 were developed by Chinese companies, but only 3 are primarily used in China. This “developed in China, used globally” phenomenon is particularly concentrated in photo and video applications. Meitu alone has five applications ranked: Photo & Video Editor, BeautyPlus, BeautyCam, Wink, and Airbrush. ByteDance is also a major player in mobile AI applications, with multiple products ranked including Doubao, Cici (general LLM assistant), Gauth (educational technology), and Hypic (photo/video editing).
Key Insight 4: Rise of “Vibe Coding” Platforms
“Vibe Coding” represents an emerging trend in generative AI applications. In the previous March 2025 ranking, this concept had just emerged with only Bolt appearing in the web ranking. In this edition, while Bolt moved to the Brink List (just missing the main ranking), Lovable and Replit made their debut in the main ranking, demonstrating rapid development in this field.
Despite sounding like a passing trend, early data shows these platforms have considerable user stickiness—enough users are staying and expanding their usage over time. Data from credit card provider Consumer Edge shows that the American user base of a leading “Vibe Coding” platform achieved over 100% revenue retention in the months following registration. This means that even accounting for lost users, the total monthly expenditure per user group continued growing, reflecting user recognition of platform value and increased usage depth.
These “Vibe Coding” platforms have also driven usage growth of other related AI products. Websites built and published through Replit and Lovable (without custom domains) have their traffic attributed to the replit.app and lovable.app domains respectively. These domains themselves have substantial traffic (lovable.app could even rank in the top 50), but still less than their developer-focused main sites.
As part of the “Vibe Coding” ecosystem, usage of some related products is growing rapidly as developers need these tools to publish their projects. While these products themselves are not AI-native companies and thus ineligible for ranking, their growth indirectly reflects the popularity of “Vibe Coding.” A典型 example is database provider Supabase, whose traffic growth has nearly synchronized with the rise of core “Vibe Coding” platforms, growing much faster in the past nine months than in previous years.
Overall, the “Vibe Coding” field offers significant growth potential and new product opportunities,预示着 further penetration of generative AI in development tools.
Key Insight 5: Analysis of Top 100 “All-Star” Lineup
Among the five editions of a16z’s top 50 web rankings, 14 companies have never missed an appearance, earning them the “All-Star” designation. These companies represent the backbone of AI consumer applications, witnessing generative AI technology’s evolution from emergence to maturity.
These all-star companies cover multiple segments of AI consumer applications, truly reflecting the full picture of consumer behavior in AI:
- 🍄
General assistants: ChatGPT, Perplexity, Poe - 🍄
AI companions: Character AI - 🍄
Image generation: Midjourney, Leonardo - 🍄
Image and video editing: Veed, Cutout - 🍄
Voice generation: Eleven Labs - 🍄
Productivity tools: Photoroom, Gamma, Quillbot - 🍄
Model hosting: Civitai, HuggingFace
When the first ranking was released nearly two years ago, the industry wondered whether all leading AI consumer companies would choose to train their own foundation models. Now we have an answer—in these 14 “All-Star” companies, 5 have self-developed models, 7 use APIs or open-source models from other companies, and 2 are model aggregation platforms. This distribution shows that in generative AI, success isn’t limited to companies with self-developed models; different technical paths can find their market positions.
Interestingly, despite the ranking’s growing globalization, these 14 stalwart companies all come from five countries: the United States, United Kingdom (Eleven Labs, Veed), Australia (Leonardo), China (Cutout Pro), and France (Photoroom, HuggingFace). This geographical distribution reflects the main concentrations of global AI innovation.
In terms of financing, all but two all-star companies have received venture capital investment. Among them, Midjourney is known for being bootstrapped, while Cutout Pro has also not raised funding, demonstrating the viability of different business models in AI.
Excluding the first edition, five additional companies would qualify for “All-Star” status, reflecting their recent strong performance. These companies also represent various AI consumer application types: Claude and DeepAI (general assistants), JanitorAI (AI companions), Pixelcut (image editing), and Suno (music generation).
In the future, as technology advances and market demands change, this list will likely continue evolving, with new star companies emerging and some existing stars potentially facing challenges.
Frequently Asked Questions
What are generative AI consumer applications?
Generative AI consumer applications refer to artificial intelligence applications providing services to ordinary consumers. These applications use generative AI technology (AI capable of creating new content rather than just analyzing existing content) to help users complete various tasks such as content creation, information processing, and creative design.
How is this ranking determined?
This ranking consists of two parts: the top 50 AI-native web products are ranked based on monthly unique visitor data from Similarweb; the top 50 AI-native mobile applications are ranked based on monthly active user (MAUs) data from Sensor Tower. Only AI-native applications—products designed with AI as their core from the beginning—are considered for ranking.
Why do so many Chinese AI applications appear in the rankings?
Chinese AI applications perform prominently in the rankings for two main reasons: first, China has a huge user base that domestic AI applications can serve; second, Chinese companies have demonstrated strong technical capabilities and innovation in certain AI segments such as video generation and image editing, with products serving not only domestic users but also successfully entering global markets.
What does “Vibe Coding” mean?
“Vibe Coding” refers to a programming approach that uses AI tools to enable developers to create applications more intuitively and quickly. This approach lowers programming barriers, allowing developers to focus more on creativity and functionality implementation rather than overly concerning themselves with underlying technical details. The rise of platforms like Lovable and Replit reflects the popularity of this trend.
What types of AI applications are most popular?
According to the ranking, the most popular AI application types include general large language model assistants (such as ChatGPT, Gemini), AI companion applications, image generation and editing tools, voice generation tools, and productivity enhancement tools. These applications directly address users’ practical needs in daily work and life.
Why don’t some large tech companies’ AI features appear in the rankings?
The ranking focuses on AI-native applications—products with AI as their core competitive advantage from the beginning. Products not originally AI-native but later adding AI features, such as Canva and Notion, are not included despite offering generative AI functionality.
What are the development trends for generative AI applications?
Several clear trends emerge from this ranking: ecosystem stabilization with increasing user loyalty to leading applications; comprehensive AI efforts by tech giants like Google; growing global influence of Chinese AI applications; emergence of new development methods like “Vibe Coding”; and diversification and specialization of AI application types.
Conclusion
A16z’s fifth edition of the generative AI consumer application ranking presents an increasingly mature and vibrant AI application ecosystem. From Google’s strong performance to the global rise of Chinese AI applications, from intense competition among general assistants to the emergence of “Vibe Coding,” each change reflects the dynamic balance between technological innovation and market demand.
As generative AI technology continues advancing and application scenarios expand, we can expect future AI applications to become more intelligent, personalized, and deeply integrated into our daily lives and work. For users, this means more high-quality tool options; for developers and enterprises, it represents tremendous innovation opportunities and market potential.
In this rapidly changing field, maintaining awareness and understanding of industry trends is crucial for both ordinary users and industry practitioners. We look forward to seeing what new changes and trends the next ranking will present, and how these innovations will further shape our interaction with technology.
