Among the 77 AI short‑form dramas that have recently emerged in China, only one broke even within six months. The remaining 98.7 % ran at a loss.
Strange, isn’t it? Thanks to AI, production costs have fallen to one‑tenth of those for live‑action. The per‑minute cost is now between 160,000 and 240,000 KRW, yet almost all of them still failed.
The numbers make the story clearer. On a single platform, Douyin (抖音, Dǒuyīn), 220,000 new titles were released this year, collectively generating over 500 billion views. Yet only 1,055 of those titles—just 0.48 %—surpassed 100 million views. The rest merely shared the 500 billion views as background noise. In the end, the short‑form platform Hongguo removed tens of thousands of low‑quality AI dramas in bulk. It’s the curation side, not the creation side, that bears the cost.
1. The Misconception That “More Production Wins”
I don’t see this as a hiccup unique to the Chinese content industry. It’s a case study that foreshadows the trap every organization that adopts AI will eventually face.
Many companies pitch AI adoption like this: “Now we can generate ten times more reports, copy, and code.” That’s not false, but the follow‑up sentence is omitted. Being able to produce ten times more also means our competitors can do the same. The Chinese short‑form market ran that experiment first. When production costs hit rock bottom, the market didn’t see a cliff—it saw a flood. In that deluge, the chance of any single title standing out fell to just 0.48 %. The colder arithmetic here is that as production costs approach zero, competitive advantage also approaches zero.
2. South Korea Is Still at a Different Stage
Of course, it would be premature to copy the domestic situation wholesale. On the Korean short‑drama platform 비글루, AI accounts for roughly 30 % of new titles, so the flood isn’t as severe as on Douyin. Production houses like KT 스튜디오지니 are still in the phase of building dedicated AI video teams. Still, I expect this trend to spill beyond the content sector. Even within Korean firms, the mantra “produce data faster and in greater volume with AI” is already being touted as a success metric. The problem is that neighboring teams and competitors can achieve the same speed with the same tools. Once volume becomes the norm, volume ceases to be a weapon.
3. So Where Should We Invest Money and Time?
When I consult on the ground, I stress three points. First, don’t mistake volume metrics for performance. Measure not “how many pieces were produced” but “how many were actually used.” Second, someone must take on the curation role, as Hongguo did. The more AI spews out drafts, the more critical the judgment of what to publish and what to discard becomes—a bottleneck and a weapon. Third, that judgment cannot be bought with tools; it’s a standard that only those who have long‑standing insight into a company’s