China's AI Drama Flood Meets Europe's New Label Law

89 of the Top 100 Dramas Were Not Filmed

In May 2026, 89 of the 100 highest-ranked animated short dramas on Douyin, ByteDance's short-video platform in China, were made without filming a single actor, according to data from the analytics firm DataEye reported by the Financial Times. A year earlier, fewer than 10 of the top 100 fit that description.

The volume behind that share is what makes it a market event rather than a curiosity. China released roughly 128,000 short dramas in the first quarter of 2026 alone, more than three times the total for all of 2025, and the Financial Times put the AI-generated share of that output at about 95 percent. Two years earlier, a single live-action short drama cost upward of 1 million yuan, roughly 137,000 dollars, to produce. By early 2026, studios using AI video-generation tools such as ByteDance's own Seedance 2.0 were finishing a comparable series for 50,000 to 100,000 yuan, with the cheapest full runs landing at 30,000 to 40,000 yuan.

The table below sets out that cost collapse in the terms Chinese studios themselves reported, with a dollar conversion added for comparison.

None of this is a side project for the companies involved. Jiangyou Culture, one of the studios that scaled fastest into the format, grew to roughly 1,000 employees and about 1 billion yuan, 137 million dollars, in annual revenue built almost entirely on the new cost structure. Kuaishou's Kling AI video-generation tool, a direct input to the boom, reported annualized revenue passing 300 million dollars in January 2026, up from about 240 million dollars a month earlier.

The Actors Who Lost the Work

The Financial Times spoke with actors living the demand side of that cost curve. Liu Yuzhan, 29, said he used to book four new roles a month; by mid-2026 he was picking up food-delivery shifts to cover the gap between acting jobs. Other actors described day rates for lead roles falling to about 300 yuan, roughly 45 dollars, for parts that used to pay several thousand yuan a day.

A separate account collected by the outlet HelloChinaTech put a number on the schedule collapse behind that pay drop: the actor Li Wenhao worked 50 consecutive days on set in 2023 and only 6 days in the whole of March 2026, as the casting calls he received fell from more than 20 a day to roughly one every two to three days.

The pay collapse is not evenly distributed. The technicians who operate the AI generation pipelines, sometimes called card-pulling technicians inside the industry, are reported to earn 2,000 to 3,000 yuan, 275 to 410 dollars, a month, below what even a diminished acting career still pays. Audience data suggests most of the flood is commercially disposable rather than a hit-driven business: of the 127,800 AI dramas in circulation by February 2026, only about 0.117 percent crossed 100 million views, and only 96 of the roughly 60,000 AI dramas launched in 2025 hit that benchmark at all. Daily ad spend on AI dramas across Douyin still reached 70 million yuan, about 9.6 million dollars, in March 2026, overtaking live-action ad spend on the platform for the first time, a model industry participants call traffic arbitrage: cheap enough to produce that even a near-total flop still turns a margin once paid promotion is added.

Beijing Cleaned Up After the Market Broke

Beijing's response arrived after the collapse in actor income and the flood of titles were already underway, not before it. Within a single week in April 2026, four separate actions landed. On 1 April, AI-generated short dramas were brought into China's national content-registration system for the first time. On 2 April, the China Broadcasting Association, an industry body, issued a rule prohibiting the unauthorized use of a performer's likeness or voice print, responding directly to a wave of AI dramas that had used lookalike characters resembling real actors including Jackson Yee, Xiao Zhan, Yang Zi, and Dilraba Dilmurat without their consent; two dramas using a Jackson Yee lookalike had already collected more than 100 million combined views before removal. On 3 April, the Cyberspace Administration of China published draft rules called the Digital Virtual Person Information Service Management Measures, and ByteDance itself announced a 200 million yuan, about 27.5 million dollars, fund earmarked for live-action production, its own acknowledgment that the platform's AI-drama economics needed a counterweight. By 6 April, the platform Hongguo had audited 15,000 AI-drama works and penalized 670 of them, under updated review standards that took effect the following day.

Every one of those four actions is a response to a market that had already reorganized itself around the cheapest possible production method. None of them caps how much AI-generated video a platform can carry, and none sets a minimum rate for the human performers whose booking calendars collapsed first.

Europe Wrote the Label Before the Flood

Europe legislated the disclosure half of this problem before any equivalent flood reached a platform operating under its jurisdiction. Article 50 of the EU AI Act became directly applicable on 2 August 2026, after the European Commission finalized its guidelines and endorsed an accompanying Code of Practice on 20 July 2026. From that date, any provider or deployer offering an AI system that generates synthetic audio, image, video, or text to users in the EU must mark that output in a machine-readable format and make it detectable by verification tools, regardless of whether anyone intended to deceive a viewer. Content that merely looks or sounds like a real person triggers the duty even without deceptive intent. A breach can draw a fine of up to 15 million euro, roughly 12.8 million pounds at current exchange rates, or 3 percent of a company's worldwide annual turnover, whichever is larger.

ByteDance is not a bystander to that rule. TikTok, the version of Douyin's own underlying technology that operates across the EU and UK, draws on the same video-generation research the company uses domestically, and any AI-generated content it serves to an EU user now falls under the Article 50 marking duty the moment that content is synthetic and detectable as such.

The genuinely non-obvious part is what the label does not fix. Article 50 requires disclosure, not a floor on how much AI content a platform may carry and not a minimum rate for the humans a cheaper alternative displaces. China's own actor-pay collapse happened while every one of those AI dramas was, in principle, legally producible; the industry body's rule against unauthorized likeness use addresses a narrower harm than the wholesale replacement of booked acting work with 200-yuan-per-minute production. A European short-drama market that reached the same cost floor tomorrow would be fully Article 50-compliant, every clip correctly labeled as synthetic, and the labor economics underneath it could still collapse exactly as they did in China. Labeling answers who made the content. It says nothing about who no longer gets paid to.

We do this for everyone trying to keep up with what technology is doing to our lives. The people who build it, and the people it happens to. The Servola Journal exists so that what we learn belongs to all of them.

Nobody pays us for this. No ads, no paywall, free to everyone. We just believe that understanding what's happening to all of us shouldn't depend on who can afford to pay for it.

If it gave you something today, tell us to keep going. Follow us, leave a like, or write a positive comment. We read every one, and they are what keeps us going.

Read next: Qwen's 3 Billion Downloads Hide a Compliance Gap  |  Anthropic's Wellbeing Grants Repeat OpenAI's Gap