AI Short Dramas Hit 1,500 Titles a Day: No Survivors at a 0.4% Hit Rate

"The entire short drama industry is in chaos right now — everyone is anxious and lost."

This was the sentence Lei Feng Network heard most often over the past month while talking with a number of AI short drama practitioners.

A workflow that was just figured out last month needs to be recalculated this month because the platform changed its revenue-sharing rules. Just when you've finally grasped how to write prompts, the AI video generation tool gets updated and the experience becomes worthless. The client's final payment hasn't arrived yet, but the production cost for the next drama has already been fronted.

The industry moves at an almost insane pace, but no one can articulate the endpoint or the direction. Right now, AI short dramas are launching at over 1,500 titles a day, with production costs compressed from hundreds of thousands or millions for live-action dramas down to under 50,000 yuan, yet teams keep pouring into the track. But almost every interviewee sends the same signal: except for the platforms, almost no one is actually making money.

A Drama's Life or Death: Just Three Days

AI short dramas are hard to understand through the logic of traditional content businesses — they're more like a high-speed probability gamble.

Short drama screenwriter Zhu Xiaoran had her work rhythm completely rewritten after quitting a big tech company. In the live-action short drama era, spending a month polishing a script was the norm; after the AI short drama explosion, "the leadership felt that with AI tools, scriptwriting efficiency should rise too." At her busiest point in the first half of the year, she delivered four scripts in one month, and after scripts were handed to the production contractor, finished films could be produced within days.

Output rose quickly, but quality soon hit a ceiling. In Zhu Xiaoran's view, a purely human-polished script can reach 90 or even 100 points, while AI-assisted scripts mostly top out at 80. "The boss knows this too, but he'd rather have four 80-point works than one 100-point work."

"This industry is essentially betting on probability through volume. We've calculated that the hit rate for AI short dramas is about 0.4%, and the rest is basically ineffective output."

The internal logic of this business model is: if you spend 6 million yuan on ten dramas and they all flop, as long as the eleventh makes 7 million and covers all losses, the model works. Live-action short dramas can be shot for 300,000 to 400,000 yuan if made crudely, while premium dramas cost 1 to 1.5 million, whereas the total investment for one AI short drama can be handled for 50,000 yuan. In terms of splits, screenwriters get a few thousand to 10,000 yuan, production gets 20,000 to 40,000 yuan, and costs across the whole chain are squeezed extremely low.

But producing episodes is only the beginning — getting exposure is the life-or-death threshold. "After a drama launches, if it doesn't pick up within three days, it's basically dead; there's no long tail." Zhu Xiaoran admits the most concentrated exposure is during the first three days of traffic buying. If it doesn't take off within three days, everyone immediately gives up on traffic buying and switches to a new drama.

Short drama director Jin Nuo, who has nine years of film and television experience, adds that short dramas rely heavily on paid traffic, and cases of pure organic virality are almost as likely as winning the lottery. AI short dramas also hit traffic bottlenecks — one short drama platform's traffic pool has a nickname, the "traffic black box." At 40 million in popularity, you're pushed to the next tier; at 50 million, you're pushed again. But AI animated dramas can't rise above 40 million.

Beyond traffic, revenue is also shrinking. Under most platform models, production-side income is divided into a guaranteed minimum paid by the platform for the script and a view-based revenue share after launch. Jin Nuo admits that in the first half of the year, China's AI short drama market could still be profitable, but now that platforms have adjusted revenue-sharing rules, self-produced projects are basically losing money. Hongguo was the first to lower the baseline revenue-sharing coefficients for AI animated dramas: the coefficient for AI realistic human dramas dropped from 60 to 40, and for 3D animated dramas from 50 to 40, while the universal guaranteed minimum for AI realistic human scripts from small and mid-sized teams was canceled. It then removed the series-drama scale coefficient in Quality AI Drama Production Campaign 2.0, narrowing the path to profit through batch-produced IP sequels.

"Only the very top of the pyramid can get meaningful revenue share. Most projects can't even recover costs."

Prompt Jockeys Earning 3,000 a Month, AI Directors Tested Within a Week

The outside world generally believes AI has lowered the barrier to entry in the short drama industry, but from the perspective of frontline practitioners, it has instead reshuffled the capabilities required of workers.

After AI entered production, a new role emerged — the prompt jockey (literally "card-draw specialist"). Screenwriters only complete the script, but AI video generation tools can't understand scripts, so someone needs to break every shot down into prompt descriptions the tool can understand. For example, if the script says "he looks at the other person," the prompt jockey has to expand it into a detailed description with facial expression and demeanor, then feed it to the AI to repeatedly generate images until a qualified shot is drawn.

This is an indispensable link in the chain and requires a lot of manpower, yet within the industry it counts as "black labor." "The boss thinks there's no technical content in it — it's just translating human language for the AI — so wages are kept very low, but KPIs are extremely heavy."

Even with the same card-drawing work and the same set of AI generation tools, different people produce wildly different results. Jin Nuo believes AI is like an accelerator, enormously amplifying the gaps in practitioners' talent and ability. "An experienced person can get a shot in two or three tries, while someone with a high rework rate needs seven or eight, and that cost difference can exceed double."

She also offers a counterintuitive judgment: the AI era actually needs professional directors even more. "AI can't be fooled. If your shots and aesthetic sense are poor, then they're poor, and the generated shots won't lie. AI only executes; it can't guess what people are thinking. In today's market, an unprofessional director won't last a week on a team."

In the live-action shooting stage, a director of ordinary ability could rely on a full crew of cinematography, lighting, and assistant directors as backup. But in the AI film and television era, the director is more like a large-scale project manager, controlling the entire production process personally, needing to understand the whole drama's visual tone and camera blocking, and also translate professional terminology into language that generation artists without film backgrounds can understand. "There's a saying in our industry that writing prompts for Seedance 2.5 is like writing the Yongle Encyclopedia — you have to specify the camera movement, image, and lighting second by second. The tools are stronger, but the threshold to master them is higher too."

Prompt jockey borrows a term from the mobile game industry. In anime-style games, "drawing cards" refers to players spending resources to randomly obtain characters or items, where input and return are not linearly correlated. AI video generation carries the same randomness: even with completely identical prompts, the model's output differs each time in composition, motion, and expression, requiring repeated generation and one-by-one comparison before a usable shot can be obtained. So the essence of a prompt jockey's work is using human labor repetitions to hedge against the model's randomness, and their output efficiency depends on familiarity with tool characteristics, precision in image description, and the aesthetic experience to judge which version is usable. This is also why the same set of tools can produce double the cost difference in different hands — rework count directly equals computing power and labor-hour consumption.

The article mentions Seedance, Jianying, Hongguo, and Juliang Qianchuan, all part of the ByteDance system. Seedance is ByteDance's video generation model, responsible for converting prompts into video clips; Jianying handles editing and post-production; Hongguo is ByteDance's free short drama app, serving as the outlet for content distribution and revenue sharing; and Juliang Qianchuan is the advertising platform for the Douyin ecosystem, through which producers buy traffic. Together these four form a closed loop from production to monetization, meaning producers lack alternatives at all four stages — model, tool, channel, and traffic buying — and their bargaining space is structurally compressed. This is precisely the technical and commercial foundation for "the biggest share of the profits ultimately flowing to the platform."

The Closer to Content, the Farther from Money

Talent determines who can stay, but the distribution of interests across the industry chain determines who can make money.

AIGC director Yue Qiaochu was among China's earliest entrepreneurs in AI film and television. The studio he founded, "Zero O'Clock Culture," has developed three business lines — TVC advertising, AI content production, and virtual idols. His AI short film Hope for the Galaxy was selected for screening at the Cannes Film Festival, and Lifelong System was shortlisted for the Busan International Film Festival's ACMF. Even with such a resume, he still couldn't escape the industry's structural dilemma, which he sums up in one sentence: the closer to content, the farther from money.

"The biggest share of the profits ultimately flows to the platform." Yue Qiaochu gives an example: video generation uses Seedance, editing uses Jianying, dramas go to Hongguo, and traffic buying relies on Juliang Qianchuan. The entire chain sits within the ByteDance ecosystem, and only the platform is guaranteed to profit. Production studios' annual revenue ceiling is mostly a few million yuan, and after deducting computing power, rent, and labor, little profit remains, with no guarantee of payment collection either.

Platforms keep raising the content threshold, while small and mid-sized teams must also bear the pressure of fronting costs. Once drama data underperforms expectations, payment cycles and advances can easily crush a company. Manpower limits constrain studio scale, making it hard to take on large volumes of projects simultaneously, and it's difficult to achieve both quality and revenue scale.

The perspectives of AI short drama clients and production contractors are similarly split. "Capable creative teams produce pretty good content, but from the boss's perspective, if that content can't make money, they won't buy it. Nowadays, plenty of people teaching others how to enter the industry and selling courses have already achieved financial freedom."

Yue Qiaochu's team is currently seeking a way out: "If you become an influential KOL or personal IP in the AI circle, your bargaining power rises, and quotes can reach 100,000, 200,000, even over a million yuan — that is, raising the price per unit of time." But he remains clear-eyed that doing B2B delivery is a certain business, and keeping the company alive is the first priority.

Is AI Short Drama Going Overseas a Good Business?

When the domestic AI short drama hit rate is squeezed to 0.4%, going overseas seems to have become the industry's collective choice, but the trial-and-error cost in overseas markets is no lower than at home.

In the view of Zhao Lei, who works for a major short drama company in South China focusing on European and American markets, European and American audiences have a stronger willingness to pay than domestic ones, which is related to the long-running paid subscription models of local platforms like YouTube. However, a strong willingness to pay does not equal easier money. Multiple industry reports point out that the payback rate of overseas AI short drama projects is extremely low, with media buying and traffic投放 accounting for more than half of total costs, and the North American customer acquisition cost per user having risen to $5.28.

High customer acquisition cost is only the first layer; overseas content barriers are also tough. Zhao Lei reveals that in overseas AI short dramas, three lines absolutely cannot be crossed: racial discrimination, minor protection, and religion. Script vetting usually has two stages: first, a domestic chief editor familiar with overseas markets reviews the first ten episodes, and if passed, work continues; after completion, an overseas chief editor reviews and checks for sensitive points.

After dodging censorship red lines, there is still a gap in content preferences. European and American audiences prefer themes like werewolves, vampires, dragons, and wizards, but it's not enough to simply wrap domestic content in a thematic coat. Domestic narrative pacing can be slow-burn, while the payoff points of overseas content must be front-loaded, and the hardest part is finding emotional common ground across cultural differences. Zhao Lei gives an example: "European and American audiences tend to favor powerful elderly protagonists," such as a drama about a 75-year-old man released from prison, abandoned by a gang family, who builds a gang from scratch, which received a good response. This may be related to the Western preference for individual heroism.

In terms of revenue structure, there isn't much difference between domestic and overseas markets — both involve rating-based revenue sharing after launch according to traffic buying, popularity, or earnings. The barrier to going overseas is not low, and the profit margin left for producers is similarly limited. For most players, it's still a situation where the platform eats meat and the producers drink soup.

Customer Acquisition Cost (CAC) refers to the average advertising spend required to acquire one paying or registered user, and for North American AI short dramas it has risen to $5.28. The corresponding metric is Lifetime Value (LTV), the total revenue a user contributes over their entire retention period. Whether the overseas short drama business model works depends on whether LTV is significantly higher than CAC. Because media buying and traffic投放 account for more than half of total costs, and overseas users' paid habits for short dramas have not yet fully developed, continuously rising CAC will quickly swallow profit margins. This is also the core reason multiple reports point to extremely low payback rates for overseas projects — not that no one pays, but that customer acquisition costs rise faster than payment growth.

In an All-Hands FOMO Market, the Disorderly Short Drama Jungle

This industry runs so fast that before the rules are properly established, the costs are already coming due.

Industry insiders reveal that in the first half of the year, a leading short drama company believed it had figured out its AI production process and launched large-scale layoffs, only to cut core execution staff and directly damage its business. "Now all bosses share a consensus that AI must be deeply integrated into the entire production process, but every company does it differently, and few have truly thought it through. Everyone is running blindfolded."

Zhu Xiaoran mentions that the boss of a major short drama company insisted the team develop an in-house AI model. "The employees actually couldn't do it at all, but they didn't dare tell the boss they couldn't finish, and in the end, when it couldn't be hidden anymore, it also ended in mass layoffs."

Many cross-industry teams with technical backgrounds who don't understand the content industry enter blindly, and most end up burning money. DataEye statistics show that in Q1 2026, active AI short drama production contractors plummeted from 1,216 to 698, a sharp 42% drop in a single quarter.

Plagiarism is also rampant. "Plagiarism is too common in the industry. Usually after a hit drama comes out, imitators rush in." Zhao Lei explains that even a one-to-one copy won't be sued as long as it doesn't become popular — if it doesn't get popular, there's no revenue or traffic data, and the copied party has no evidence even if they want to sue. He also shares firsthand observations: many small companies require screenwriters to submit the first ten episodes as a test when hiring. "After the script is submitted, they say it passed, but halfway through writing they say they're not doing it anymore, and then you find out the company used AI to wash the script."

When talking about the industry, Zhu Xiaoran raises one point: "The reason short dramas can become popular is fundamentally that audiences always have a demand for good stories." But at a pace of 1,500 titles a day, "good" is giving way to "fast." Yet the demand for good stories won't disappear. Who the next 0.4% will fall to — the industry itself is still waiting for the answer.

One-line Summary

AI short dramas hit 1,500 titles a day with only a 0.4% hit rate, leaving the whole industry in FOMO while only platforms reliably profit.

Paragraph Summary

AI short dramas now launch at over 1,500 titles a day with production costs compressed under 50,000 yuan, yet the whole industry is gripped by collective anxiety. The hit rate is only about 0.4%, and a drama is abandoned if it fails to gain traction within three days, making paid traffic the life-or-death threshold. In the chain, prompt jockeys are in high demand but low-paid with heavy KPIs, while professional directors are scarcer; platforms keep tightening revenue-sharing rules, with Hongguo cutting coefficients and canceling guarantees, leaving producers "closer to content, farther from money." Going overseas is seen as a way out, but North American CAC has reached $5.28 with extremely low payback rates. In Q1 2026, active contractors plunged from 1,216 to 698, plagiarism is rampant, and blind spending continues as the industry waits in disorder for the next 0.4%.