
DouBao App, which has been free for years, is now selling memberships.
Recently, many users have noticed that DouBao's page in the Apple App Store has been updated with a description of its four-tier pricing structure for paid subscriptions, ranging from a minimum of 68 yuan per month to a maximum of 500 yuan.
However, DouBao did not specify the differences in benefits across versions, only mentioning that it will“include more value-added services.”。
DouBao's official response stated that DouBao continues to offer free services. On top of the free service, DouBao is also exploring the introduction of additional value-added services, with relevant plan details still in the testing phase.
According to a report by 'First Finance,' insiders revealed that DouBao’s paid features will mainly focus oncomplex tasks and productivity scenarios, such as PPT generation, data analysis, and video production.
The above-mentioned individuals stated that as model capabilities continue to improve, the product is already able to handle an increasing number of complex and high-value tasks. However, such tasks require more computational power and longer reasoning times, so DouBao plans to launch a paid service to meet the needs of these complex scenarios.
It's not hard to see that after DouBao introduces its paid membership, light users who only chat or look up information won't be affected.
However, heavy users who frequently use DouBao for drawing, making videos, creating music, designing PowerPoint presentations, or developing applications will likely need to recharge in order to use the services without restrictions. Additionally, DouBao's latest large models, especially multimodal models like Seedance 2.0, may also be prioritized for paid users.
Differentiating between free users and paid members is quite common within AI circles.
Domestic companies like Kimi and MiniMax, as well as international ones like OpenAI, Google, and Anthropic, have all launched their own paid membership systems. Among them, membership fees account for over half of OpenAI's total revenue.
As everyone knows,large-scale AI models have gradually moved beyond the stage of burning money without considering costs.Regardless of whether they are large, medium, or small companies, there is an increasing emphasis on commercial returns from AI businesses; even if they aren't profitable, efforts are being made to minimize losses.
Especially for OpenAI and Anthropic, both companies plan to go public this year and must improve their financial performance, leading to wave after wave of commercialization efforts around AI apps.
Now, DouBao has taken its first step in charging users. This once again signals Zhang Yiming's intention to 'account' for AI and make it profitable.
However, DouBao has primarily focused on consumer-facing scenarios in the long term, while its application in business-oriented work scenarios remains relatively weak. The willingness of C-end consumers to pay is generally low; to make such users willingly spend money, DouBao may find it difficult with its current product and feature offerings.If Zhang Yiming wants DouBao to generate revenue, the first step is to help DouBao identify its true 'selling point.'
The best selling point for an AI app is productivity, especially AI programming capabilities. Competitors like ChatGPT and Claude, which possess this ability, have successfully sold premium memberships; a contrasting example is Grok, which so far can only chat, create images, and produce videos, remaining more of a 'toy.'
Similarly, if DouBao remains confined to general lifestyle scenarios and fails to integrate into professional office environments, then its potential for selling memberships will likely be as limited as Grok’s.
At the same time, this is also an 'identity recognition' experiment.
Through charging, ByteDance will have the opportunity to filter out high-value, heavy users from DouBao's 300 million monthly active users.and tailor more refined paid packages specifically for them in the future, thereby amplifying the commercial value of its existing user base through meticulous operations.
A
Compared with competitors, DouBao’s pricing standards are not considered low:
——Basic Version: Free to use, meeting daily needs.
——Standard Version: Monthly subscription for 68 yuan, yearly subscription for 688 yuan.
——Enhanced Version: Monthly subscription for 200 yuan, yearly subscription for 2,048 yuan.
——Professional Version: Monthly subscription for 500 yuan, yearly subscription for 5,088 yuan.
In the domestic market,DouBao's main competitors, Qwen and Yuanbao, are both free to use.Qwen even publicly stated in November last year, 'Ordinary users can use it anytime for free. We currently have no plans to charge.' DeepSeek, which has the second-highest user base after DouBao, also does not offer a paid membership.

The AI Six Dragons present a different scenario.
Relatively weaker financial resources have forced the Six Dragons to start commercialization earlier, with paid memberships being one of the primary approaches.Their monthly fees range from 30 to 60 yuan, while advanced memberships can reach several hundred yuan.
For example, Kimi offers four pricing tiers, with the basic version at 49 yuan per month for auto-renewal and the highest tier priced at 699 yuan. The differences between tiers mainly lie in Agent quotas, whether they support multi-task parallel processing by Agents, access to AI programming, support for professional databases, and functionality like 'raising shrimp'.
Foreign AI companies generally charge higher fees.
Since 2023, OpenAI has been selling ChatGPT memberships, initially offering only the Plus version at 20 US dollars per month and the Pro version at 200 US dollars. Later, a 100 US dollars per month plan was added.
Not long ago, it also launched ChatGPT Go, priced at 8 US dollars per month in the United States and around 5 US dollars per month in countries like India.
Another AI giant, Claude, aligns its pricing structure with OpenAI's, offering three tiers: 20, 100, and 200 US dollars.
In comparison, DouBao, which starts at 68 yuan (approximately 10 US dollars) per month, is quite reasonably priced.
B
However, a slightly higher or lower price is not the real obstacle.The most critical question DouBao needs to answer is why users should subscribe to its membership?
When selling memberships, both domestic and international AI companies almost unanimously emphasize that paid versions are suitable for work scenarios.
Kimi’s membership page lists use cases including Office document processing, in-depth research, website deployment, as well as support for financial, economic, and academic data from platforms such as Tonghuashun and Tianyancha. It specifically highlights 'access to Kimi Code', where higher membership tiers provide greater quotas for this AI programming feature.
Moreover, advanced members can utilize Agent clusters to enhance productivity in scenarios like search, writing, and batch processing through parallel multi-Agent operations.
OpenAI also focuses primarily on productivity, especially in AI programming.
According to the official introduction, ChatGPT Plus can try out advanced features such as Codex and in-depth research. The Pro version, priced at 100 USD per month, offers five times (limited-time ten times) the usage of Codex, while the 200 USD version reaches up to twenty times (limited-time twenty-five times).
As for Anthropic and Google, their AI App pricing strategies and core service scenarios are not significantly different.Charging tens or hundreds of dollars monthly, they specifically serve the working class — particularly software engineers.
This is precisely DouBao's weak point.
Since its launch, DouBao’s functionality matrix has grown increasingly extensive, but it mostly revolves around general lifestyle scenarios. For example, core functions like the Chatbot and AI creation, which carries the genetic makeup of Douyin content.
Users typically open DouBao to ask a question, look up information, write some text, or simply 'tease' the AI. Some people use DouBao for drawing, making videos, or even composing a song, mainly for daily entertainment purposes.
In contrast,Work-related scenarios within the DouBao app cannot be said to be nonexistent, but they are not mainstream.Currently, the productivity features it incorporates, such as PPT generation and small application development, are relatively 'superficial,' falling under the category of light office tasks.
The resulting challenge is that C-end users who come for entertainment are probably not strongly inclined to purchase memberships.
After all, the free version of DouBao can already meet most daily needs. Paid memberships might offer access to more powerful models from ByteDance’s suite, increased usage of multimodal models, and even early access to state-of-the-art (SOTA) products, but this is not an essential need for the average person.
DouBao focuses on lifestyle scenarios and lacks productivity-related features, differing fromByteDance's AI product strategy.
AI programming represents the core productivity scenario for AI today. ByteDance recognized this early on but did not integrate it into the DouBao app. Instead, they launched TRAE and Kouzi, which focus on AI-integrated development environments and vibe coding, respectively.
To date, TRAE and Kouzi have developed well and gained popularity among the coder community. It can be said that ByteDance has achieved its strategic goals set earlier.
However, a side effect of this is thatDouBao has long been 'lopsided,' excelling in lifestyle scenarios but lagging significantly in productivity scenarios.
C
AI is ultimately meant to get work done, and this is especially true for paid AI services.
An AI app lacking office functionality will find it difficult to sell memberships and may only resort to unconventional strategies or offering large volumes at low prices to compensate.
For instance, Grok, under Musk's umbrella, charges a fee comparable to ChatGPT and Claude but falls significantly short in AI programming and other office-related capabilities.
To appear cost-effective, Grok offers a much higher volume of multimodal usage than its competitors. Paid users can generate hundreds of videos daily through Grok, with virtually no limit on image generation.
Even so, whether considering total user numbers or paid memberships, Grok cannot compete with the leading players.
DouBao is now facing similar challenges as Grok.Although DouBao surpasses Grok significantly on the model side, it still cannot serve as a cyber worker akin to ChatGPT Pro.
Despite the considerable difficulty, DouBao has embarked on the path of charging users. The reason is that earning revenue from B-end clients isn't easy for DouBao, necessitating C-end income to subsidize operations.
DouBao not only has an app but also offers models, smartphones, and hardware businesses such as earphones and recording devices.

Though it operates various business lines, nearly all of DouBao’s current revenue comes from the B-end market, primarily by selling tokens via API and MaaS to developers and enterprises. This segment is managed by Volcano Engine.
DouBao has not disclosed its revenue figures. However, in January this year, the Financial Times cited data from market research firm IDC, stating that Volcano Engine has become China’s second-largest provider of AI infrastructure and software, capturing 13% of the market share. In the first half of 2025, ByteDance’s revenue from China’s AI cloud services market reached $390 million.
Compared to quarterly revenues of approximately $50 billion,ByteDance generates less than 1% of its revenue from this segment, which is almost negligible.Moreover, only a portion of this revenue is related to the DouBao large model, making the latter's actual revenue scale even smaller.
As for hardware like smartphones, these projects are still in their early stages of development, with more emphasis on testing the waters than generating income, and thus cannot contribute meaningful revenue.
In this situation, the simplest and most direct way for DouBao to quickly increase revenue would be to sell memberships.
According to a report by market research firm QuestMobile, as of March this year, DouBao had 345 million monthly active users, with an average of 54.8 uses per user per month.Even if only 1% of the monthly active users pay, DouBao could earn over 200 million yuan more each month, and nearly 3 billion yuan more annually.
Overseas peers have already proven the feasibility of this approach.
As early as 2024, subscription fees for ChatGPT (including both individual and enterprise subscriptions) contributed about 60% of OpenAI’s revenue. In 2025, as OpenAI vigorously developed advertising and other businesses, this figure dropped to 55%, still accounting for half of its revenue.
OpenAI places great importance on membership fees, continuously lowering the threshold to attract consumers. At the end of April this year, OpenAI launched the cheaper ChatGPT Go, aiming to reach 112 million subscribers within the year; by 2030, this number is expected to increase to 306 million.
According to reports by media outlets such as The Information,The proportion of paid users for ChatGPT is approximately 5% of the weekly active users.If DouBao could reach this level, its monthly revenue might exceed 1 billion yuan.
Of course, DouBao is targeted at the domestic market, where Chinese consumers are accustomed to free services and are particularly cautious about paid services. The conversion rate for DouBao's paid services is expected to be significantly lower than that of overseas companies.
Even so, DouBao’s large user base is still sufficient to generate a considerable membership fee. With B-end business still under development, DouBao steadily earning subscription income from the C-end remains an efficient path to commercialization.
D
DouBao's membership model can not only make money but alsosegment the traffic pool, distinguishing heavy users from regular users.
Although DouBao has many users, it hasn't solved a key issue until now: how to differentiate between various types of users.
For social media or video platforms, all they need to do is analyze user behavior and interaction patterns to easily identify individual profiles, then 'prescribe the right medicine' by marketing products and services to them—a process known as 'monetizing traffic.'
However, DouBao is an AI app mainly focused on chatbot-driven general lifestyle scenarios. Different people using DouBao won’t vary drastically in their usage patterns, limiting the granularity of differentiation.
It’s time for DouBao to start differentiating its users:Users with unclear profiles, even if more numerous, are far less valuable than precise users.

One piece of evidence is that third-party statistics show each monthly active user of OpenAI contributes $25 in revenue; for Anthropic, which specializes in AI programming, this figure is $211—eight times higher than OpenAI's.
At the corporate level, while OpenAI has far more users than Anthropic, its annual recurring revenue falls short of the latter’s, and there are even signs its valuation may be surpassed.
One of the most effective ways to break through this situation is by selling memberships.
Users willing to pay for a membership must have stable, high-frequency needs for advanced functions, far exceeding those of regular users.These needs are often related to office scenarios. By capturing this group, you capture the most valuable segment of the entire user pool.
This also partly explains why OpenAI would rather lower its monthly fee in order to sell more memberships. Only by differentiating between free users and paying customers can OpenAI operate comfortably in terms of commercialization.
Another interesting aspect is that OpenAI ties membership fees to advertisements—users opting for the cheapest $8 monthly plan might see ads in ChatGPT, similar to the free version.
While this is certainly one of OpenAI's revenue-generating methods, it can also be understood as follows:The 'price-sensitive consumers' filtered out by the cheaper plans are precisely the target audience for these ads.
Correspondingly, those users willing to spend hundreds of dollars each month on recharges are the key group that OpenAI needs to focus on maintaining.
DouBao is also making a similar choice.
DouBao's membership sales represent Zhang Yiming’s new financial move.
Like other major tech companies, DouBao’s costs mainly include infrastructure such as chips, cloud computing power, and data centers, as well as technology research and development like large model training and app development, daily operations such as large model inference and app maintenance, and expenditures on marketing, staff salaries, and bonuses.
Reports indicate that ByteDance’s capital expenditure for 2025 will exceed 150 billion yuan (approximately 21.6 billion US dollars), with the majority concentrated in the AI sector. Additionally, this year ByteDance has allocated 85 billion yuan (approximately 12.2 billion US dollars) for AI chip procurement. Compared to revenues in the hundreds of millions, investments at the tens of billions of dollars level are two orders of magnitude higher. And this is only part of ByteDance’s AI investment.
Whether it is infrastructure construction such as AI chips, or the R&D, training, and inference of foundational models, as well as user acquisition and maintenance for apps, all require long-term investment. Especially since other major players are competing for dominant entry points, although DouBao currently leads in user numbers, it cannot afford to be complacent or reduce its investment.
However, alongside these investments,ByteDance also hopes that DouBao can generate more revenue, or at least demonstrate its potential to make money.
Previously, DouBao’s commercialization efforts on the consumer side were mainly focused on hardware. Most users have not yet developed habits of shopping or listening to music within the DouBao app, the market performance of AI hardware has been lackluster, and DouBao’s mobile assistant has failed to break through the 'encirclement' by competitors.
In the January article 'It’s Time for Zhang Yiming to Reckon with DouBao,' LetterList proposed thatDouBao might as well be more 'aggressive', following the example of peers like OpenAI, and straightforwardly monetize memberships and traffic.
Now, this idea has officially come to fruition.
If in 2025 DouBao is busy competing with DeepSeek, Yuanbao and others, not wanting to set membership barriers that could turn away some users, then by early 2026, DouBao's dominance in the AI App track will allow it to confidently start charging membership fees.
However, as previously mentioned,users are more willing to pay for productivity scenarios, which is precisely DouBao’s weak point and one that cannot be solved easily in the short term.
Selling memberships can generate revenue and differentiate users, offering numerous benefits; however, as things stand, DouBao needs more than its existing product and feature matrix to carve out a piece of this pie. This AI App, built on a Chatbot foundation, faces the key challenge of transitioning into dual 'life + work' scenarios if it wants to monetize memberships successfully. $ByteDance (FT0001)$$ByteDance Ecosystem (LIST1266.HK)$$AI (LIST0535.SH)$$Artificial Intelligence (LIST2136.US)$$Artificial Intelligence (LIST23586.HK)$
Risk Disclaimer: The above content only represents the author's view. It does not represent any position or investment advice of Futu. Futu makes no representation or warranty.Read more
Comments
to post a comment
