AI Short Drama Hit the Hot Search, and Live-Action Shoots Fell 70%

On October 7, a hashtag translating roughly to "after watching AI short dramas, I have to say live-action ones are finished" climbed to second place on Weibo's trending list, with a heat score just over 15 million. Someone wrote that they had assumed machines could not replace real actors, watched four AI dramas, and changed their mind. Emotion aside, the trending page carried numbers that are hard to argue with: of the top 20 titles on the Hongguo drama chart, only one used real actors. The other 19 were AI-generated.

Younger viewers no longer care much whether a show was made by people or by a model. They scroll, they keep watching if it holds them, they swipe away in two seconds if it does not. That reflex says more about what has happened than any industry report.

An order-of-magnitude gap in cost, and a bigger one in capacity

Put the prices side by side and the outline of the change is clear. An AI short drama costs roughly a tenth of what a live-action one does. A live-action production that used to run 300,000 to 400,000 yuan now lands for tens of thousands. Three people and five days can turn out 80 episodes. Among short dramas launched in the first quarter, more than 95 percent were AI-made, and a new one appeared roughly every 36 seconds.

That is not only cheaper. The structure of production has changed. A live-action crew is bound by actor schedules, locations, weather, and post-production timelines. An AI pipeline has none of those limits. It can test ten premises a day and pour resources into whichever one performs. For a platform, that is predictable, batchable, takedown-ready supply.

{{INLINE1}}

What got squeezed was the studio floor

The cost lands on specific people. Live-action shoots fell about 75 percent year over year, and vacant space at the Hengdian studio complex is close to 80 percent. After platforms dropped their minimum revenue guarantees, a batch of high-cost, low-return projects simply stopped, and capital turned toward AI. Many mid-tier actors found they could not get work, not because of their craft but because their names were never on the schedule.

People in the industry like to call this metabolism. The word is not wrong, but it dodges a fact: the people being metabolised do not automatically grow new skills. The short-drama labour chain is long. Extras, lighting, grips, costume and set crews, editors. Each link can hold hundreds of workers, and their re-employment speed is nowhere near the speed of model releases.

Thrill is the engine, and formula is the ceiling

AI short dramas win by serving the appetite of the lower-tier market precisely. The fantasy, cultivation, and war scenes a live-action crew cannot afford? An AI produces them at will, with stable-looking leads and polished costumes, and no risk of a star scandal or a fee hike. Audiences never wanted celebrities. They wanted feeling. If a show cranks the conflict to maximum in the first 30 seconds, retention takes care of itself.

The problem sits in the same place. Formulaic scripts produce volume, not hits. More than 90 percent of AI short-drama producers are losing money, and only a handful in the industry are actually profitable. Viewers are also getting more sensitive to the "AI smell." Stiff expressions, empty eyes, and flat emotions come up again and again in the comments. The distance between "you cannot tell it is AI" and "I can tell instantly" is only a few interchangeable shows.

The real dividing line is whether it feels alive

Funnily enough, the highest praise a viewer can give an AI short drama is that it does not feel like a machine. Another trending topic in October asked what happened when a grandmother was placed into a film by AI and left speechless. The focus was not on realism but on an emotion that could not be put into words. That tells you audiences are not rejecting AI. They are rejecting cheapness.

Live-action short drama actually walked a detour over the past few years. After the boom in 2024, everyone wanted to make prestige work, supply of low-end drama thinned, and a slice of the audience drifted away. AI filled exactly that gap: low barrier, short cycle, plenty of volume. So rather than saying AI eliminated live-action drama, it is fairer to say live-action drama gave up a piece of ground it no longer wanted.

What happens next depends on two things. One is whether the feeling of being alive can be patched in with technology. Performance detail and the eyes are the hardest part, but the pace of model progress over the past year should stop anyone from concluding too early. The other is whether compliance and labelling can keep up. AI content is increasingly required to disclose itself, and platforms are tightening their revenue rules. Once supply turns from scarce to surplus, survival stops being about whether you can make it, and starts being about whether anyone remembers it.

Look at it from another angle and the real change is the cost structure of trial and error. A premise used to be worth shooting or not based on experience, because shooting was expensive. Now a failed experiment costs a few thousand yuan, so a producer can back a dozen directions at once and double down on whichever one shows data. That pushes content production closer to how internet products work: move fast, kill quickly, concentrate resources on what is already validated. It does not guarantee a hit. It turns a gamble into a calculation.

Hot searches fade and charts change hands.