Old‑school long‑form dramas are proving harder for AI to kill — and China’s platform structure helps explain why

Human craft still sparks what AI can’t

A growing chorus in China’s cultural press argues that long-form television — the slow, knotty dramas that build mood and meaning over dozens of episodes — resists the AI era because it depends on “high‑loss” authenticity and counterintuitive human choices. The point is simple and sharp: generative models predict probability and therefore converge on averages. They can synthesize surfaces quickly. They struggle to produce the strange, brittle combinations that make scenes feel genuinely lived-in. Recent domestic examples often cited are the period romance 《纯真年代的爱情》 and the crime drama 《除恶》 — shows where location work, odd casting choices and single, risky beats (an off‑type actor playing against stereotype, a murder staged in a lotus field) deliver emergent moments that feel irreducibly human. Can an algorithm that optimizes for the common case create those misfired, combustible moments? The answer looks increasingly like no.

Directors and crews matter. Teng Huatao (滕华涛) — known for urban marriage and romance dramas such as 《蜗居》 and 《双面胶》 — reportedly surprised viewers and critics by pivoting toward darker, more unpredictable themes, a move that lessens model‑predictability but heightens human resonance. On productions like the Hubei‑set period piece, producers rebuilt factory courtyards, stitched together real buildings and staged quotidian scenes that accumulate atmosphere over time. That patient, iterative labor is not a data point; it’s a craft. Short formats and AIGC can knock out polished textures in minutes. They cannot, at present, manufacture the slow burn of ensemble workmanship — the collective “handmade” that many Chinese viewers still prize.

Platform alignment, FOMO and why China’s situation differs from Hollywood

The tension is not only artistic. It has been reported that China’s streaming platforms double as both the gatekeepers of distribution and the vendors of AI tooling — iQiyi (爱奇艺) sits within Baidu (百度)’s ecosystem and its AI efforts link to Wenxin Yiyan (文心一言); Tencent Video (腾讯视频) is part of Tencent (腾讯) and tied to that group’s models; Youku (优酷) sits under Alibaba (阿里) and its Tongyi Qianwen (通义千问). When the people who finance and distribute your show also sell the AI that promises to “lower costs and boost efficiency,” incentives skew toward scaleable, formulaic output. In the U.S., studios and legacy media have pushed back — the New York Times sued OpenAI, and Hollywood distributors wield copyright leverage — but in China those institutional checks are weaker because the tech and media interests overlap.

What follows is industry FOMO: executives pressured to deploy AI to cut budgets and accelerate schedules, even if the cultural return is dubious. Some viral AI projects have won attention more for savvy promotion than storytelling, it has been reported, underscoring the difference between novelty and durable art. The likely near‑term outcome is painful and partial: AI will siphon out low‑effort content — the “water” — making business models leaner, but it will struggle to dislodge the genuinely non‑standard works that depend on particular human experiences, risky moves and time to emerge. For China’s TV industry the diagnosis is internal: not that long‑form is obsolete, but that years of algorithmic shortcutting have hollowed out the craft that creates true emergence. Art still bets on idiosyncrasy. AI can clone the average.