Key Points Summary
In the second quarter of 2026, the most notable change in Zhihu's financial report was a significant reversal in its revenue structure: paid content and IP operations accounted for 61.7%, surpassing advertising (28.8%) for the first time. This indicates that Zhihu has shifted from a business model that relies on selling advertising based on user attention to one where content is treated as an asset that can be reused multiple times—for example, selling story copyrights to short-form TV series producers, expert knowledge to AI companies, and community content for use by AI systems. However, these new businesses are still in their infancy, and the revenue they generate has not yet made up for the decline in advertising revenue, so it remains uncertain how much growth they can achieve in the future.
I. Why Did Advertising Fall from a Main Source to a Secondary One? Three Reasons Why Zhihu Couldn't Succeed with the Traffic-Based Business Model
The essence of advertising is to sell user attention: the longer users stay on a platform and the more frequently they interact, the more advertisers are willing to pay. However, Zhihu is not well-suited for this approach due to three main issues:
1. Misaligned User Attention Patterns: Zhihu users come to the platform to find answers and leave quickly (with an average daily stay of 41 minutes), whereas Bilibili users spend more time consuming content (with an average daily stay of 113 minutes). Advertisers need users to engage deeply with the platform, which Zhihu cannot provide, resulting in lower ad prices.
2. Lack of Scale: In previous years, Zhihu invested heavily in acquiring users, but it failed to retain them, leading to unsustainable revenue. Bilibili, with 370 million monthly active users, can continue to grow its advertising revenue; Zhihu does not have that scale and cannot play the same game.
3. Proactive Removal of Low-Quality Advertising: To maintain a clean community, Zhihu removed more than 40% of low-quality commercial content, which temporarily reduced revenue. However, this was a deliberate choice, not a forced one.
Now, Zhihu focuses on advertising in industries that match its user base, such as technology and automotive, and treats it as a supplementary revenue stream rather than a primary one.
II. The New Business Model: Selling the Same Content Multiple Times, from “One-Time Consumption” to “Recurring Revenue”
Zhihu has transformed the 993 million pieces of content it has accumulated over the past fifteen years from being discarded after use into assets that can be sold multiple times. The main ways it does this include:
1. Copyright Sales for Short-Form Media: For example, the story “Wrong Marriage, Happy Ending” by author Jiu Sen was first sold as a subscription on Zhihu, then its copyright was sold to a short-form TV series producer, and it has also been sold for use in manga and films. In the second quarter, copyright revenue increased by 671%, with over 1,000 works having their copyrights sold.
2. Expert Data for AI Companies: Zhihu's SoTALab network connects with over 200,000 experts, including doctors and lawyers, to provide training data for large models (such as designing reasoning questions and evaluating models). While ordinary data annotation costs a few dollars per piece, expert data can fetch hundreds of dollars per piece—this is a significant difference in value. Similar to Reddit overseas, Zhihu is selling more valuable expert data.
3. Content for AI Use: Zhihu is making its community content available for developers to use in AI applications (such as search engines and intelligent assistants). The number of AI developers using this content increased by 50% in the second quarter.
III. New Businesses Are Promising, but Revenue Is Still Limited
Despite the promising new directions, Zhihu's revenue is still small:
- Low Base of Copyright Revenue: Last year, the total revenue from copyright, vocational education, and other products was only 366 million yuan, while advertising revenue decreased by 800 million yuan in the same period. In the second quarter, the highest-paid author received only 300,000 yuan from copyright sales, which is far from sufficient.
- Declining Member Base: The average number of monthly subscribed members in the first half of the year was 13.1 million, 600,000 fewer than last year. Membership fees are a significant source of revenue for authors, but this pool is shrinking.
- Declining Gross Margin: To support content creation, Zhihu's content costs increased by 10.4%, reducing its gross margin from 62.5% to 57%. This means using profits to invest in the future.
- Early Stage of AI Data Business: Compared to Reddit, which generates $43 million in annual data licensing revenue, Zhihu's expert data revenue is not yet separated as a separate category and is included in “other income,” making it a minor part of its total revenue.
IV. Zhihu's Advantages and Uncertainties: How Big Can the New Business Become?
Zhihu has several strengths, but the future is uncertain:
- Financial Strength: It has 4.4 billion yuan in cash, which will help it through the initial stages of developing these new businesses.
- Early Advantage in Short-Form Content: Zhihu launched a paid subscription model for selected content in 2021 and has a stock of over 200,000 licensed works.
- Early AI Investment: It began working on AI-related projects in 2024, and its AI-related services have seen 37% growth this year.
However, there are also many uncertainties:
- Potential Scale of New Revenue: The current revenue from these new businesses is limited. Can Zhihu increase the number of times content is sold from a few times to many times?
- Competitive Threats: Other platforms, such as Tomato Novel (which produces 6,700 short-form works per year) and Yuewen (which is investing in AI-generated content ecosystems), pose competitive threats. Can Zhihu maintain its advantage in the content licensing market?
- Industry Risks: The majority of AI-related content is in the form of short-form videos, and the market for real-person-starring short-form content is shrinking. Could Zhihu's model be disrupted? There are also no unified standards for AI training data licenses, which could lead to legal issues.
In Conclusion
Zhihu has demonstrated that content can be sold multiple times, but it has not yet proven that this can generate substantial profits. The key to its future success lies in how often content can be sold. Whether Zhihu can expand this business significantly will determine its long-term viability.