Doubao will start charging fees officially in late June and accelerate the integration with Douyin e-commerce | Exclusive
In early May, the news that Doubao would launch a paid subscription service sparked extensive discussions in the market. The updated paid subscription plan in the Apple App Store shows that Doubao will introduce four tiers of pricing: Basic, Standard, Enhanced, and Professional. The corresponding monthly fees are: free, 68 yuan, 200 yuan, and 500 yuan; the annual fees are: free, 688 yuan, 2048 yuan, and 5088 yuan.
Subsequently, the official responded, "Doubao always provides free services. On top of the free services, we are also exploring the introduction of more value - added content to meet the diverse needs of different users."
According to exclusive information from 36Kr, the above is just a warm - up for Doubao's commercialization. In the coming quarters, Doubao will continue to promote the implementation of commercialization. A person familiar with the matter revealed that Doubao is expected to officially launch paid content in late June and update relevant functions at the Force Conference held during the same period. The reason for choosing this time node is that it will take about a month for the PC and mobile versions to complete the adaptation and transformation of basic functions and the charging system.
According to 36Kr, if things go smoothly, Doubao will further integrate e - commerce functions to improve the paid scenarios in the third quarter and attract traffic to the Douyin Mall through subsidies. It will enter the operation phase in the fourth quarter. These actions are all preparations for long - term commercial returns in 2027 and beyond. Therefore, in 2026, Doubao will not use the penetration rate of paid users as an evaluation indicator.
In October 2025, Doubao initially connected with Douyin e - commerce. Users can obtain product information and shopping link recommendations through conversations with Doubao. According to the aforementioned person familiar with the matter, ByteDance once tested Doubao's hard - advertising capabilities on a small scale at that time, but due to user experience issues, the large - scale roll - out was later suspended. Currently, Doubao adopts an interactive mode where it prioritizes responding to user questions and then inserts horizontal product recommendations in the answers.
The aforementioned person familiar with the matter said that since April 2026, users' acceptance of AI shopping recommendations has increased compared to the second half of last year, and the conversion rate of clicking on product cards and viewing the detail pages has reached over 3%. On the other hand, the user growth of Doubao has slowed down. According to 36Kr, part of the slowdown is also a result of ByteDance's intentional decision - the high cost of computing power inference has led ByteDance to reduce its promotion and advertising investment in Doubao in the new year.
The common rhythm of mobile Internet products is to burn money to acquire users first and then engage in monetization, but the commercialization of AI products is obviously more urgent. ByteDance is no exception.
01
Is Subscription a Good Choice?
According to a report by Bloomberg on May 27, ByteDance is considering investing up to $70 billion in capital expenditure to promote AI development. Less than a month ago, on May 9, the South China Morning Post reported, citing a person familiar with the matter, that ByteDance had raised its AI infrastructure budget from 160 billion yuan last year to 200 billion yuan (about $30 billion). The continuous increase in investment shows that ByteDance is betting on AI with unprecedented intensity.
The global AI arms race is still heating up, but how to make money has become a real problem for all companies.
Previously, 36Kr reported that although ByteDance's overseas revenue increased by nearly 50% in 2025, due to the huge investment in AI business, the overall net profit decreased by more than 70% (the official later clarified that this decrease was mainly due to international accounting standards). With the continuous investment of huge amounts of money, global technology companies are finding it difficult to stay in the stage of only talking about technological ideals without considering commercial returns.
In the current market, the three most commonly mentioned paths for AI commercialization are subscription models, advertising monetization, and To B model services. Among them, the subscription model is the first commercialization scenario to be explored.
Looking back at the overseas market, leading AI products actually started exploring paid models very early. After its launch in November 2022, ChatGPT only took about two months to launch the ChatGPT Plus subscription service at $20 per month in February 2023. Subsequently, Anthropic also launched Claude Pro in September 2023, also priced at $20 per month. In contrast, domestic Chatbot products have remained in the free stage for a long time, still following the competitive strategy of "burning subsidies to seize user volume".
As the first domestic Chatbot product planning to switch to a subscription model, Doubao does have a deeper user base. According to the global ranking of the AICPB product list, the monthly active users of the Doubao App in April 2026 were approximately 336 million, with a month - on - month growth of 1.51%, ranking second globally after ChatGPT.
The problem is that Doubao's ability to quickly become a leading Chatbot product is not mainly due to the intelligence of its model, but more to the usability at the product level. Coupled with the powerful traffic delivery ability of the Douyin ecosystem, Doubao has captured the attention of users in the sinking market first.
In fact, domestic users do not have strong loyalty to a single AI product. According to a report by LatePost, the average daily usage time per user of Doubao is currently 10 minutes. Without significant technological barriers and with competitors still offering free services, opening up subscriptions may lead to user loss.
However, regardless of the product itself, the subscription model is not necessarily the best path for AI commercialization.
Even in the Western market where users have a stronger willingness to pay, the revenue conversion that subscriptions can bring is still limited. To increase the number of paid users, ChatGPT launched the cheaper ChatGPT Go in 2026, which only costs $8 per month. This increased the number of its paid users from 47 million at the end of last year to 55 million in the first quarter. But according to The Information, the average weekly active users of ChatGPT in the first quarter were approximately 905 million. In other words, even for the Chatbot product with the largest global MAU, after three years of implementing the subscription model, the paid conversion rate is only 6.1%.
The difficulty in increasing the number of paid users lies in the dilemma of the AI subscription model itself.
On the one hand, in a highly competitive and homogeneous product market, no model company dares to completely cancel free services. For most casual users, the free version is sufficient to meet their basic needs for daily Q&A and information queries. On the other hand, when tokens are still expensive, to use AI to complete more complex tasks, the current maximum price can reach $200 per month (about 1400 yuan). And only a small number of people are willing to pay for it in the long term.
More importantly, even with such a high price, AI companies may not make money. In June 2025, Sam Altman publicly stated on social platform X that OpenAI was losing money on the $200 - per - month ChatGPT Pro subscription. This leads to a paradox - if the price is too high, users are reluctant to pay; if the price is too low, it is difficult to cover the computing power cost.
Therefore, C - end subscriptions are more of a "losing - money marketing strategy" at present. To truly achieve commercial returns, technology companies still need to explore more commercialization scenarios. For example, OpenAI started to gradually launch and promote advertising business in ChatGPT in February this year.
02
How to Implement AI Commercialization
After a long period of setbacks, technology giants seem to have finally found the best path to commercialization - selling model capabilities to enterprises. Enterprises do not need to train their own models. They can directly call the AI model capabilities developed by technology companies through API access and pay according to the token usage.
Compared with individual users with limited willingness to pay, enterprises have the ability and willingness to use these AI tools to improve internal business efficiency. More importantly, in the To B model, enterprises usually pay continuously according to the token usage, rather than subscribing on a monthly basis (as mentioned before, the monthly subscription price for ordinary users usually cannot cover the computing power cost). For model manufacturers, this is obviously a profitable business.
Anthropic's profit growth has fully demonstrated this. According to a report by The Wall Street Journal on May 21, Anthropic's revenue in the second quarter is expected to more than double to $10.9 billion, and it may achieve an operating profit of $559 million. In an industry still in the stage of large - scale subsidy burning, Anthropic is the first large - model company approaching profitability. Its main source of income is API calls from enterprises and startups. As of October 2025, Anthropic had more than 300,000 enterprise customers, accounting for about 80% of its total revenue.
Domestic Internet giants have also successively deployed and made efforts in the ToB market. Different from AI startups like Anthropic, Internet giants have cloud computing capabilities: on the one hand, they monetize through model APIs. This form of encapsulating Model A as a cloud service for external output is MaaS (Model as a Service); on the other hand, the computing power consumption caused by model calls may drive the growth of their own cloud businesses.
In this new AI cloud competition, ByteDance's Volcengine has the fastest growth. On May 7, data released by the International Data Corporation (IDC) showed that in 2025, the large - model call volume of Chinese enterprises on public clouds increased by 16 times compared with the previous year. Nearly half of the call volume occurred on the Volcengine platform, with a market share of 49.5%, ranking first in the industry.
Alibaba also launched the "Bailian Campaign" at the end of November 2025, trying to catch up in the AI cloud market share. According to Alibaba's financial report data for the fourth fiscal quarter of 2026, the number of customers on Alibaba's "Bailian Platform" (Alibaba Cloud's MaaS platform) increased by 8 times year - on - year. This also shows that Alibaba is rapidly strengthening its AI To B market capabilities.
In addition to the obvious benefits brought by the To B market, ByteDance is also actively using AI capabilities to empower its main business.
Take ByteDance's ace AI model, Seedance 2.0, as an example. Its video - generation ability has further promoted the creation boom in the short - drama market. A person in the short - drama industry told 36Kr that as the threshold for video creation continues to decrease, a large number of new players have entered the industry. They have diverse backgrounds, coming from the gaming, trendy toy, and even traditional manufacturing industries. In addition, the reduction in the production cost of a single short drama also allows production companies to produce more dramas with the same total budget.
The surge in the number of creators and the reduction in production costs have both increased the supply of AI short dramas. At present, the most core way for short dramas to obtain views is still advertising on the Douyin side. A person in charge of short - drama advertising revealed that the peak daily consumption of AI short dramas in ByteDance's Juleiang Engine system is about 120 million yuan.
Wang Xiaoshu, the founder of Jiashu Technology, which is responsible for short - drama distribution, told 36Kr that his company's daily advertising consumption was only a few hundred thousand yuan last year, but this year it has reached several million yuan, a nearly ten - fold increase. At the same time, the ROI of short - drama producers on Juleiang Engine is usually only between 1.03 and 1.07. This means that for every 1 million yuan invested in advertising, a profit of about 30,000 to 70,000 yuan can be obtained.
It can be seen that behind the continuous expansion and prosperity of the short - drama industry, it is still mainly the platform itself that benefits stably.
AI + e - commerce shopping may become ByteDance's next area of focus. On March 30, 2026, according to Jiemian News, Doubao's shopping payment function is undergoing a gray - scale test. This function allows users to directly place orders and complete payments for products within the Doubao APP without jumping to the Douyin Mall.
Doubao is not the first AI product to attempt to complete the AI shopping loop. On May 11 this year, Alibaba Tongyi Qianwen was officially connected to Taobao, completing the AI shopping function. In September last year, ChatGPT also launched the "Instant Checkout" function, trying to guide users to complete orders and payments directly in the AI dialog box. However, in March this year, OpenAI cancelled this function. The reason is that while it is easy to complete the technical path, it is difficult to integrate the entire e - commerce ecosystem behind it.
The advantage of Doubao and Tongyi Qianwen is that they are backed by the two mature e - commerce systems of Douyin Mall and Taobao respectively. Whether it is the product library, merchant ecosystem, payment, or fulfillment capabilities, they already exist. This also makes AI shopping seem more feasible in theory.
However, even so, AI shopping still has to face two core questions. Firstly, is AI shopping a real need? During the Spring Festival, Tongyi Qianwen invested 3 billion yuan in red envelopes to encourage users to order milk tea through AI, trying to cultivate consumers' habits of making consumption decisions through AI. 36Kr also reported on this. But from the results, most users did not form long - term retention. And compared with ordering takeout, shopping is a more complex scenario that relies more on comparison and decision - making.
Secondly, will consumers really trust AI product recommendations? As the importance of AI search and recommendation capabilities increases, some merchants have begun to influence the information retrieval and sorting of large models through methods such as GEO (Generative Engine Optimization). The GEO industry chain exposed at this year's 315 Gala also sparked extensive discussions. When AI recommendations are influenced by commercial advertising and interest relationships, whether consumers will regard them as objective suggestions or new advertising channels remains to be seen.
(Zhou Xinyu also contributed to this article)
Editors | Qiao Qian, Yang Xuan
This article is from the WeChat official account "36Kr Future Consumption", author: Xiao Sijia. Republished by 36Kr with permission.