The AI app earned $1.2 billion in 2024, nearly doubling its previous earnings. Which users have started to spend money on AI?

Half of 2025 has passed. Although the Product-Market Fit (PMF) of AI still troubles entrepreneurs, it doesn't affect the profitability of mobile AI products.

According to a report released by Sensor Tower this month, in 2024, the in-app purchase revenue of global non-game apps increased by $13.7 billion year-on-year to $69.2 billion. Among them, AI apps only accounted for $1.2 billion, but the year-on-year growth was 179%. And according to Sensor Tower's classification, productivity tools with increased revenue due to the integration of AI also performed well.

Do Photo/Video Apps Rely on Asia to Survive?

This wave of AI started at the end of 2022. 2023 was a period of wild growth, and by 2024, the entire track became more mature. The download volume of AI products reached 1.49 billion times, a year-on-year increase of 91.8%, and the revenue reached $1.2 billion, a year-on-year increase of 179.1%.

However, different from the diverse Web-based AI products, the main product forms on the app side are still ChatBot and Art Generator, which was also a previous consensus. Therefore, this interpretation still focuses on these two types of products, but more on user portraits.

From the perspective of revenue, the distribution of this category shows obvious regional characteristics. In markets such as Europe, America, and the Middle East, only a few products can enter the Top 10 of the AI product revenue list in each market. These include DaVinci and Momo under Turkish manufacturers, as well as AI Mirror developed by a Chinese team, which are basically products we have observed before.

Products in the Asia-Pacific region that entered the overall AI product revenue list

Image products | Data source: Sensor Tower 

Note: The geographical divisions in the original report are AMER (North America, South America), EMEAR (Europe, the Middle East, Africa, Russia), and APIC (Asia-Pacific region). Therefore, Asia mentioned below specifically refers to the Asia-Pacific region. However, from the perspective of user usage habits, Australia is more similar to Europe and America, so it is excluded when making this table. In addition, due to the special nature of the domestic market, products in the domestic market are not included.

And the "new faces" mainly appear in Asia. Eight Art Generators can enter the Top 10 of the revenue list in each Asian market. Besides comprehensive image generation products, some products have emerged in the segmented categories of some markets, such as face-swapping products in Taiwan, China, Japanese users' preference for anime products, and AI hair style changing in Thailand.

However, even so, according to the data in 2024, Art Generator products still mainly focus on the image modality, and there are few video products. But by mid-2025, we can start to observe that AI video generation apps such as Kling, PixVerse, and Hula have started to generate good revenue. Even Pixverse has reached a monthly active user (MAU) of 60 million, and the revenue is also in the hundreds of thousands of dollars. (We will release a relevant topic about Pixverse next week. Welcome to follow.)

ChatBot: "Highly Educated" People Choose ChatGPT, while General Users Prefer Wrapper Products

The top 5 global AI products in terms of MAU (left) and revenue (right)

Except for ChatGPT and C.AI, they are all wrapper products

The ChatBot category can be roughly divided into Chatbots with self-developed models, wrapper Chatbots, and AI companion products.

As far as we know, ChatGPT is still the well-deserved leader. It accounts for 23% of the download volume and about 40% of the revenue of all AI apps. But surprisingly, even though manufacturers such as Google, Microsoft, and Anthropic have launched their own ChatBot apps, wrapper products still firmly occupy the top positions in the global MAU and revenue lists. From the data trend, there is no obvious sign of decline for these wrapper products, and they are still doing well.

As can be seen from the above figure, AI social/companion products have a low presence in the Top 5, with only C.AI. But when expanded to the Top 10, we can also see Talke. Even in some special markets, some AI companion products can enter the Top 3 in terms of revenue.

Among them, products with a relatively large MAU such as Talkie, PolyBuzz, Xingye, Maoxiang,

hi waifu, Linky, Zhumengdao, etc. are all from domestic manufacturers

In our observation of global AI social apps in April, AI role-playing dominated, while other product forms such as virtual companions, game companions, and AI star/influencer avatars accounted for a relatively low proportion.

Here, we need to specifically mention LoveyDovey. This product is mainly a mixture of AI role-playing and AI star/influencer avatars. It simulates the whole process of two people getting to know each other, having an ambiguous relationship, and falling in love through a series of gamified designs, and attracts fan groups through the IPs of online novels, online comics, and idol stars. In 2024, LoveyDovey's revenue ranked second in the Japanese and South Korean markets, only lower than ChatGPT. (Readers can refer to the article "An AI Social App 'Abandoning the US Market' with Monthly Revenue Crushing C.ai") 

Overall, from the product side, there are no very subversive changes in this report, but in terms of user behavior, there are some more detailed insights.

First, in terms of user age and gender, currently AI ChatBot products still mainly target users who are "sensitive to new technologies and new things and have certain productivity needs". The most representative user portrait is "male users aged 25 - 34". The proportion of male users of ChatGPT and Microsoft Copilot both exceeds 70%, and the proportion of male users of other products is also around 60%. But over time, the proportion of female users of ChatBot is also gradually increasing. In mid-2024, the proportion of female users of ChatGPT increased by 4%, and that of Copilot increased by 7%.

For AI companion products led by c.ai, the users are mainly young female groups aged 18 - 24. The proportion of female users of c.ai is 71.8%. According to Diandian Data, the proportion of female users of Talkie also reaches 60%.

Besides the general user insights, SensorTower gave detailed user portraits and product preferences of four representative AI products, ChatGPT, Copilot, c.ai, and ChatOn, in the report.

Let's first look at the comparison between "ChatGPT" and "the wrapper tool ChatOn". Overall, users of ChatGPT are more "elite" and have a stronger willingness and ability for self-learning, while users of ChatOn are more inclined to use tools for efficiency improvement.

Among the preferred products of ChatGPT users, there are developer communities, anonymous workplace communities, financial apps, and even online courses and meditation apps. It can be seen that this type of users will actively obtain information, pay more attention to self-learning and life management, and actively look for diversified products to meet their own needs.

In contrast, as a "wrapper product", ChatOn attracts users who are used to using various tools for efficiency improvement. The products preferred by this type of users mainly provide "simple and convenient" value, such as AI education, drawing, video editing, and cleaning tools.

ChatOn's landing page (1), segmented function page (2),

Prompt Engineering (3),

Model and chat style adjustment page (4)

ChatOn also responds to the needs of these users in product design. Different from ChatGPT's "start chatting directly on a blank page", ChatOn provides more explicit function sections. After users send messages in these segmented functions, ChatOn will automatically perform Prompt Engineering according to the scenario.

On the other hand, besides GPT, ChatOn also provides diversified model/reply style options such as Claude and DeepSeek. In addition, there are other tools such as image generation, AI keyboard, and PDF interpretation. The more task-oriented product design and more diversified model and tool options make it more attractive.

From the perspective of C.ai's user product preferences, besides other AI companion products, there are Gacha Games, otome games, as well as Alight Motion (animation design APP), Facemoji (AI input method), etc. Overall, the user group of C.ai "is relatively emotional, likes self-expression, and has a certain desire for creation". Since C.ai is mainly driven by UGC, which is essentially a model where the official provides the model and technology, and some users produce content for other users to consume, the users' own creative desire is the key to promoting the product content iteration.

However, "a good product is not necessarily a good business". From the charts of daily active users (DAU), user stickiness, and revenue, we can see that c.ai has an extremely high user usage time of 90 minutes per day (close to the level of short - video products such as Douyin), but the revenue is relatively low. First of all, since the platform does not produce IP and relies entirely on external IPs, c.ai has no pricing power, resulting in the dilemma that although the traffic is high, it cannot make money from IPs and can only earn a small amount of subscription fees. On the other hand, the high computing power cost caused by users' frequent use makes the team unable to bear it, and finally it was acquired by Google.

From our recent observation, AI companion products are also exploring in the directions of compliance, multi-modal content, and diversification such as game companions and AI influencer avatars.

With the Arrival of AI, the Spring of Tool Products is Back, but the Big Manufacturers Still Benefit the Most

Besides native AI products, the productivity tool category is another highlight in this report. From the data, although the overall download volume has decreased by about 2%, the user usage time and in-app purchase revenue have both increased significantly. In particular, the revenue has increased by 34