AI Short Drama Melee: Are Big Tech Firms Vying for Scripts or the Next-Generation Traffic Entry Point?
In 2026, AI-generated short dramas have become one of the most popular tracks on the Chinese internet.
According to data released by QuestMobile, the net growth of total usage duration of short drama videos reached 133.29 billion hours in September 2026; the monthly active users of Hongguo Free Short Drama have exceeded 430 million. Within the Douyin Group, its user count even surpasses that of Doubao, making it the second-largest application only after Douyin.
Also in this year, the four giants ByteDance, Tencent, Kuaishou and Baidu have entered the market one after another. Some have developed independent apps, while others have built mini-program matrices, breaking down the capabilities of large video models into individual independent business modules, and each has launched supporting AI short drama creation incentive plans.
On the surface, the competition among giants focuses on three things: who develops their own app, whose revenue-sharing policy is more attractive to production teams, and who has a larger script library. But the real question that needs to be answered is another one: what the big players have been betting on all along, is it a batch of short drama scripts that can be put on the market for consumption, or the next-generation traffic entrance?
01 Four paths reveal the same underlying anxiety
The four giants have vastly different entry strategies, but their paths themselves have already laid their cards on the table.
ByteDance is currently the only player that has deployed the entire industrial chain.
The Seedream image model, Seedance video model, Jimeng AI and Jianying jointly form the production tool layer, Fanqie Novel provides content source supply, Douyin and Hongguo undertake traffic distribution, and user behavior data flows back in reverse to guide the next round of topic selection decisions. In other words, this system is basically a closed loop: AI-generated content first goes through market verification in ByteDance's own traffic pool, and the hit topics that perform well are immediately put into the next round of mass production. Hongguo's 430 million monthly active users, coupled with the 151 minutes of usage duration per capita per day, is the most direct manifestation of this closed loop in operation.
In addition, in October 2026, Volcano Engine launched another creation incentive plan — the "Seedance Premium AI Drama Incentive Plan", which is graded according to the monthly ranking of popularity on Hongguo: the top 1 to 10 works will be rewarded 200,000 yuan per piece, top 11 to 20 will be rewarded 100,000 yuan per piece, top 21 to 50 will be rewarded 50,000 yuan per piece, and the incentives are distributed in the form of 50% cash plus 50% equivalent coupons.
This plan is not a simple cash subsidy, but a carefully designed strategic initiative, with the core of taking content incentives as leverage to seize the dominant right of the AI film and television production chain. If you look closely, ByteDance's funds are not spent on acquiring new users, but on the supply side, what it buys is content density. Others calculate customer acquisition cost, while ByteDance calculates how much content needs to be consumed per unit of time. These two accounts are not the same thing at all: the former is a traffic business, and the latter is a factory business.
But the closed loop is also a wall. If topic selection is determined by backflow data, it means the platform will continuously produce those emotions that have been verified, there will be more and more hit works, and they will look more and more similar. ByteDance itself should be well aware of this, so what it really needs to look for outside is exactly new stories that have not yet been verified by data.
Tencent's strategic anxiety needs to be analyzed from its business data.
QuestMobile data shows that in June 2026, the market share of user duration of Tencent's products dropped to 29.7% from 31.9% in the same period of the previous year, falling below 30% for the first time. The urgency lies in that ByteDance's share rose from 33.3% to 40.1% in the same period.
As early as June 2025, WeChat launched an official short drama mini-program for trial operation first. In July 2026, this mini-program was renamed "Green Bubble Short Drama" and started gray-scale testing. Within one month after its launch, the number of active users reached 156 million, second only to Hongguo, making it the second largest short drama application.
In October 2026, WeChat officially launched Green Bubble Short Drama. Its incentive policy has a very special design: it no longer takes play volume as the only assessment standard, but sets the threshold that the per capita play duration is no less than 10 minutes, and new dramas that meet the standard can get 100% income incentives.
What Tencent is doing is actually a defensive offense. What it is anxious about is not how much money the short drama business itself can make, but the meaning behind the number of "user duration falling below 30%": WeChat is the last super entrance in China whose duration has not been completely occupied by short videos. If the consumption habit of short dramas is developed elsewhere, what WeChat loses is not only a piece of revenue, but also the part of content mind outside the social relationship chain.
It is precisely because of this that it dares to set the incentive threshold at "10 minutes per capita". Play volume can be brushed, but duration cannot. What Tencent wants is not a good-looking ranking list, but to prove that users can also stay in WeChat.
Kuaishou's approach is much more focused, and its core logic is only one: keep a close eye on commercial realization.
Public financial reports show that the operating revenue of Keling AI in the second quarter has exceeded 850 million yuan, an increase of more than 200% over the same period of last year, and an increase of about 31% over the previous quarter. In addition, the cumulative revenue in the first half of 2026 exceeded 1.5 billion yuan, which is already higher than the 1.1 billion yuan for the whole year of 2025.
Kuaishou takes Keling AI as its main underlying architecture. In 2026, it launched special incentive policies for high-quality dramas and series dramas, with a maximum subsidy of 200,000 yuan for a single drama. It also held AI creator competitions, with a maximum cash support of 500,000 yuan for a single project.
Among the four companies, only Kuaishou has counted AI short dramas as "revenue" rather than "strategy". This is a quite smart path: instead of competing with ByteDance for the position of content platform, it chooses to sell shovels.
But the ceiling of the tool business is determined by the customers' ability to pay.
Kuaishou subsidizes high-quality dramas and holds creation competitions. It seems to be building a content ecosystem, but in essence it is driving demand for its tool customers. Only by subsidizing buyers can it sell more shovels. How far this play can go depends on whether the downstream production contractors can still hold on.
Finally, let's talk about Baidu. Its industrial layout is relatively marginal, but its strategy is more pragmatic.
In January 2026, Baidu Intelligent Cloud launched an AIGC full-process solution for the manhua drama industry, covering all production links such as script generation, storyboard disassembly, picture generation, dubbing and soundtrack, and also took the lead in establishing the AIGC content creator alliance.
On the Baidu APP side, according to the disclosure at the Baidu Creator Conference: the scale of AI manhua drama users on the platform reached 100 million in April 2026, the creator revenue-sharing scale in that quarter increased by 103% compared with the previous quarter, and the cumulative revenue-sharing of the platform reached 2.2 billion yuan. By August, the quarterly creator revenue-sharing scale increased by another 88% month-on-month.
Baidu looks the most inconspicuous, but its position is the most clever. It basically does not participate in the competition for entrances, but provides infrastructure for all those who enter the competition. At the same time, it uses the information flow scenario of Baidu APP to carry the manhua drama form which is lighter, cheaper and more suitable for the sinking market. 100 million manhua drama users and 2.2 billion cumulative revenue-sharing show that it has found the category gap ignored by the other three companies, but its weakness is also obvious: the growth of revenue-sharing scale is highly dependent on platform subsidies and advertising inventory. Once advertisers lose patience with the conversion effect of manhua dramas, this curve will break faster than the other three.
Behind the four paths is the same judgment: AI short drama is one of the few tracks that continue to generate incremental space after the pattern of the mobile internet is solidified. However, what they are really competing for is not the same: ByteDance is competing for supply density, Tencent is competing for duration retention, Kuaishou is competing for tool orders, and Baidu is competing for category gaps. When the four things are pushed forward at the same time, they will all hit the same bottleneck: no matter which path you take, where does the available content that can start production come from?
Thus, the common problem of the industry is pushed to the next layer: the script.
02 The battle for scripts: what is being fought for is not stories, but assets that can be put into production repeatedly
Is there any real evidence for the "script determinism"?
In terms of fundamentals, most of the AI short drama hits in the first half of 2026 are adapted from online literary IPs. For example, the total play volume of "The Record of Ten Thousand Demons" across the whole network exceeds 6 billion, and "Exiled to the Border, the Convicted Wife Develops Wasteland and Raises a War God" has a play volume of more than 4.1 billion on the Hongguo platform. Both works are adapted from online literary IPs. On the contrary, original AI content without IP support basically has no living space.
As a result, the licensing fee for online literary IPs has also risen accordingly.
Data from the Ent Group report shows that S-level works cost 80,000 to 100,000 yuan per piece, popular A-level works cost 60,000 to 80,000 yuan, and ordinary works start at 30,000 yuan; from January to May 2026, the transaction volume of online literary IP licensing reached 1.8 billion yuan, an increase of 120% over last year.
But concluding that "who gets the scripts wins" is a one-sided analysis.
In traditional film and television works, the script is the creation blueprint of the work, but in the AI short drama industry, the script is more like a standardized production mold. This mold does not focus on literariness, but on whether it can be disassembled and reused: the world view should be large enough to continuously expand new plots, the conflict density should be high enough to have an emotional outbreak within 3 minutes per episode, the structure should be standardized enough to be produced in seasons and translated into multiple languages, and multiple teams can start work at the same time, and so on.
In other words, what makes hit online literary works really valuable is that they have been screened by millions of readers, verifying which emotions can make people pay repeatedly. Following this logic: if the script is really the core competitiveness, then Yuewen, Fanqie and Qimao, which have the thickest IP reserves, should be sure to win.
But the actual situation is that IP parties are expanding downstream to the production links.
In January 2026, Yuewen invested in Soy Sauce Culture, a leading AI production company, to increase investment in the creation of full-category animation content, and also specially set up the "Manhua Drama Assistant" service for manhua dramas; Qimao and Zongheng still follow the logic based on the IP library from the free reading era, and use the guaranteed minimum plus revenue-sharing model to attract production parties to cooperate.
The position of the script in the AI production process has also changed.
In the era of live-action dramas, screenwriters are the starting point of content creation. The production of AI short dramas is deduced backward from post-production: the editor first determines what pictures are needed for each segment, then goes back to disassemble the script and generate materials. The script has changed from a "blueprint" to a "material list".
At the same time, DataEye report shows that the number of active production contractors in the middle reaches of the industry has decreased from 1216 to 698, a decrease of 42%. The unit production price has been suppressed to 300 to 500 yuan per minute, and the net profit margin of OEM production is only 5% to 8%. The more standardized the production mold is, the narrower the living space for pure OEM production becomes. This also answers the question left in the previous section: how big Kuaishou's shovel selling business can be depends on how much profit the people holding the shovels have left.
Therefore, the essence of the script battle is not who has stronger creative ability, but who can turn stories into standardized assets that can be licensed, updated continuously and reused in batches, so as to maintain profit margins in a market with excess supply.
Once the content assets are created, they need to find a place to realize value, which brings us to the third battlefield.
03 The settlement rule is the key lifeline
The entrance battle in 2026 is divided into two paths.
One is independent APP.
QuestMobile data shows that Hongguo had 168 million daily active users in July, with a growth rate of 107%, which is more than the sum of the daily active users of iQiyi, Tencent Video, Youku and Mango TV; in September, its monthly active users reached 430 million, with a per capita daily usage duration of about 151 minutes. This set of figures shows that as long as the content is cheap enough and the supply is sufficient enough, short dramas can stand out from a large number of content categories and grow into an independent consumption scenario, rather than an attached column of a certain platform.
The other is ecological embedding.
Tencent's Green Bubble Short Drama does not have an independent APP, and runs directly in WeChat. It had 143 million monthly active users in August, and 89.68 million weekly active users in the first week of October, with a month-on-month growth of 44.1%. Users search, watch and share in WeChat, and do not need to leave the platform throughout the whole process.
Baidu's Wenju Hui adopts the mini-program form, packaging short dramas and manhua dramas for advertising realization. Kuaishou not only operates the short drama traffic on its main site, but also opens up an independent Xifan position.
The two paths have their own accounts. Independent APPs have relatively controllable advertising inventory, user duration and algorithm recommendation, but they are expensive to acquire users and difficult to retain users; mini-programs have low customer acquisition cost and fast social fission, but their duration and advertising spaces are limited by the host ecology, and essentially they have to pay rent to the host.
Therefore, the first layer of the entrance battle is the contest between download and click efficiency, and what determines the victory is customer acquisition cost and per capita duration, not the DAU (Daily Active Users) ranking list.
Therefore, the real key falls on the settlement rules.
For the revenue sharing of content parties, the generally accepted basic calculation method in the industry is: (newly added viewing duration) * (basic duration unit price) * (content type coefficient) * (copyright coefficient) * (incentive coefficient) * (completion and replay correction coefficient). But there are also some subtle differences and fine-tuning. For example, AI dramas have an additional "art style coefficient", which has been adjusted three times in the first half of 2026.
This calculation method means that the platform does not need to control the supply by buying out copyrights at all. As long as it adjusts the coefficients, it can determine what and how much to produce in the market. This is the real meaning of "entrance": the entrance is not a faucet, but a water price meter, and the company that sets the coefficients is the bookmaker.
Therefore, what the platform really holds in its hands are three things: what content can be launched, how much traffic it can get after launch, and in what proportion it is settled. This also explains why the four big factories seem to have very different actions on the surface, but in fact they are putting together the same puzzle.
Fighting for scripts or seizing the next generation of traffic entrances are essentially just means. Scripts are reproducible means of production, and entrances are negotiable distribution channels. What is really scarce and cannot be copied is the power to price all of this.
Whoever masters the settlement coefficients will simultaneously determine what to write in the upstream, what to shoot in the midstream, and how much to earn in the downstream. This is probably the most noteworthy point in this melee: on the surface, the four big factories are competing for