In May 2026, 89 out of the top 100 AI dramas and animated series on Douyin were AI-generated human-like dramas
DataEye reports that nearly 90% of the top 100 AI dramas on Douyin in May belong to AI-generated content.
According to statistics from DataEye, in May 2026, among the top 100 AI dramas and animated series on Douyin, 89 were AI-generated human-like dramas, contributing approximately 29.97 billion additional views, accounting for nearly 90% of the total increase on the list; the rest included 4 commentary animations, 1 comedic animation, and 3 each of 2D and 3D animations. The top show of the month, "Exchange of Spouses to Become Pets," had an increase of about 870 million views in that month.
Supply has peaked and begun to decline. In May, Douyin added approximately 39,500 AI dramas and animated series, a decrease of about 10% from approximately 44,100 in April, marking a decline for two consecutive months; the April additions were already over 13 times the number of real human short dramas added during the same period, which was about 3,248. In May, there were still 252 new dramas that surpassed 100 million views, with a success rate of about 0.64%. In the first half of the year, Douyin launched over 221,900 new dramas, with only 1,055 surpassing 100 million views, resulting in a hit rate of about 0.47%; approximately 12,000 surpassed 10 million views, accounting for 5.8%. According to the China Network Audio-Visual Association, there were about 367,000 micro short dramas across the internet in the first half of the year, with over 74% generated by artificial intelligence; out of approximately 128,000 short dramas in the first quarter, about 122,000 were AI micro short dramas, accounting for over 95%.
The production process has shifted from film crews to prompt-based pipelines. The New York Times cited industry figures stating that some companies can produce content for about $30 per minute without requiring real actors; CNN reported that a five-person team can produce 120 minutes of average quality content in about a week at a cost of about 15,000 RMB, which is less than one-tenth of the low-end costs of real actors. ByteDance's Seedance 2.0, launched in February, and its subsequent versions are used by producers for human-like dramas and animated series; ByteDance also owns Douyin, Hongguo short dramas, and a library of web literature available for adaptation. New regulations for managing micro short dramas by the National Radio and Television Administration are expected to require clear labeling of AI-generated series starting September 1. The Hongguo platform has already removed over 20,000 low-quality AI dramas in two rounds from April to May.
Profits have not kept pace with production. A DataEye representative stated that production costs have decreased by about 90%, while advertising costs have increased by over 100% year-on-year, with revenue dropping by over 50%; earnings per 10,000 views have fallen from about 30 to 100 RMB last year to about 5 to 10 RMB now. Advertising can account for about 70% of the total cost of real human-like AI dramas. Overseas, TikTok short dramas reported over $21 million in monthly revenue share, with AI short dramas accounting for over $6.5 million, a month-on-month increase of about 40%; the global market for AI short dramas is estimated to be about $100 million in 2025, with projections of about $650 million in 2026. The overall market size for domestic micro short dramas was previously estimated by DataEye to be nearly $15 billion, exceeding the box office figures for the same period.
Actors and image rights have begun to push back. Celebrities threaten to sue for unauthorized use of their likenesses, while real actors report a decrease in job offers. In cities like Hangzhou and Jiaxing, thousands of people are involved in AI drama production lines, with local governments offering subsidies of up to about 2 million RMB per drama. Following the expansion of production, there are complaints about homogenization and aesthetics: shows that start off invincible, rely on special effects, and have hollow stories are being eliminated by both platforms and users.
In terms of market mechanisms, the buyers are users on Douyin and Hongguo who pay per episode or watch ads, while the sellers are workshops and distributors that update daily using models like Seedance. Transactions are driven by cost reductions from generation tools, rather than improvements in drama quality. Funding comes from user payments and information flow advertising into production and user acquisition; as production becomes cheaper, budgets shift towards traffic auctions. The beneficiaries are the platforms, model providers, and a few distributors with over 100 million views, while the ones under pressure are real drama crews, diluted mid-tier workshops, and the illusion of revenue from views—over 90% of new dramas do not recoup their user acquisition costs.
Source: Public Information
ABAB AI Insight
China's short drama industry has completed a supply revolution: transforming actors, locations, and post-production into model calls. DataEye's statistic of 89 out of 100 is not an aesthetic victory; it reflects the dominance of human-like AI content in Douyin's hot list. Seedance operates within ByteDance's own distribution empire, with models, web literature copyrights, and two player platforms forming a closed loop, which Hollywood lacks.
The capital path involves leveraging after cost reductions to buy traffic. Production costs have dropped by 90%, advertising costs have doubled, and earnings per 10,000 views have fallen by over 50%, determining that most shows are expendable. The removal of 20,000 dramas by Hongguo indicates that the platform is already using compliance and quotas to reclaim the open period. The mandatory labeling starting September 1 will change "human-like" from a default trust to an explicit synthetic label, requiring another assessment of user acquisition efficiency. AI dramas have already captured a visible share of overseas revenue, with NetShort and DramaWave ranking high on international charts, while domestic competition has been exported into another set of paywalls.
The benchmark is the online novel factories and short video information flow advertising of the 2010s: replicating the same emotional hooks at the lowest cost, with a hit rate of less than 1%. The industry phase has shifted from expansion to control—production has declined for two consecutive months, with live daily updates giving way to a few titles that can surpass user acquisition thresholds.
Structural changes involve technology replacing labor and a shift in pricing power. What is being replaced are on-screen labor and small to medium-sized crews; pricing power has shifted from production costs to platform recommendations and user acquisition auctions. The mechanism is: generative models compress marginal costs to near zero, and recommendation systems sort by completion and payment, thus competition has shifted from "can it be produced" to "can the first million views be bought." The 89 human-like dramas in the top 100 are faces chosen by algorithms, not people chosen by drama.