A viewer watching the AI-produced drama “The Peach Blossom Pond” on Anhui TV Image generated by AI

In late July, regular viewers of Anhui TV noticed something unusual about the dramas airing during prime time. On screen, an ink-wash landscape of the Jiangnan region slowly unfurled—the rippling waters of Peach Blossom Pond glistening in the light—while a small caption remained fixed at the bottom of the frame: “AI Production.” “The Peach Blossom Pond,” a drama centered on intangible cultural heritage and generated entirely by AI, had made its television debut.

With no starring actors and no on-location shooting, every frame was generated by AI—a first for a domestic satellite TV channel. Long confined to mobile and online platforms, AI-produced short dramas were now making their first tentative move onto mainstream television. Yet as viewers began weighing in, a heated debate over technology, art, and aesthetic standards quickly spread from the comment sections into the broader public sphere.

From small screen to big screen

The arrival of AI-produced short dramas in satellite TV prime time may appear to be a tentative step toward greater media convergence, but it also brings two very different production models and evaluation standards into direct contact. Over the past two years, AI short dramas have proliferated on mobile internet platforms, fueled by extremely low production costs and short production cycles. Most follow a formulaic, traffic-driven approach—heavy on conflict, rapid-fire plot twists, and cliffhangers designed to capture viewers’ attention within seconds. Genres such as CEO romance and martial-arts fantasy are mass-produced and endlessly replicated, resulting in severe content homogenization. Until now, such productions have largely remained outside mainstream audiovisual channels, often dismissed as “fast-food content.”

After “The Peach Blossom Pond” premiered, the Weibo topic “Anhui TV airs AI-produced drama” quickly attracted nearly 20 million views, with reactions sharply divided. Supporters hailed the broadcast as an experiment worth encouraging. On social media, many viewers pointed to the particular advantages of AI-based production—virtual actors have no scheduling conflicts, while niche cultural tourism themes and traditional crafts can be depicted without the high costs or logistical difficulties of live-action production. Others, taking a more tolerant stance toward new technologies, argued that the industry remains in its infancy and that its current audiovisual shortcomings can still be improved.

The criticism, however, was more direct and scathing. Many viewers took aim at the drama’s glaring audiovisual flaws and its undeniable lack of a convincing human touch. Some noted that faces intended to convey subtle emotions lacked the micro-expressions of real human performances, while frequent technical missteps proved even more immersion-breaking. Among the trending comments, one particularly popular post read, “Anhui TV sure knows how to save money”—a teasing remark that reflected a widespread perception that AI short dramas are little more than a cost-cutting compromise by the networks.

According to Liu Yongchang, a professor at the School of Journalism and Communication at Nanjing Normal University, while AI can open up forms of visual imagination beyond the reach of live-action filming, particularly in grand or speculative settings, and expand the boundaries of storytelling, its limitations are equally significant. At present, AI-generated imagery struggles not only to produce coherent, detail-rich depictions of real-world scenes, but also to replicate the layered, complex performances that human actors create through individual personality, psychological depth, and specific dramatic contexts. These expressive shortcomings, rooted in a lack of humanistic depth, are unlikely to be fundamentally resolved even if producers succeed in improving visual quality and narrative pacing for television broadcast.

A deeper divide lies in the vastly different review standards applied by satellite TV and online platforms. Li Jinsha, an associate professor at the School of Television at Communication University of China, noted that mainstream audiovisual platforms prioritize works with cultural communication value. This requires them to rigorously guard against “AI hallucinations,” such as the arbitrary accumulation of cultural symbols that generative AI can produce. They also require complete, traceable copyright documentation throughout the production chain, as well as audiovisual quality that meets television broadcast standards. By contrast, online platforms continue to rely heavily on traffic data and short-term completion rates, while allowing far greater latitude in genre and considerably more tolerance for flaws in execution.

Reshaped industry chain

The rise of AI short dramas on mainstream television reflects a broader technological shift that is reshaping the film and television industry chain. The most visible changes are in production efficiency and cost. AI’s batch-generation model can significantly compress production cycles, while eliminating the high costs of location shooting and large-scale set construction. Yet these technological gains have also brought growing pains as the industry adjusts. Over the past two years, the number of live-action short drama projects has continued to decline, while traditional film and television professionals—actors, cinematographers, and others—have faced mounting pressure from potential job displacement. Anxiety over whether AI will replace human workers has consequently spread across the industry.

The impact of AI on traditional production roles is real, but simple replacement is unlikely to define the industry’s evolution. He Tianping, an associate professor at the School of Journalism and Communication at Renmin University of China, noted that a clear stratification has already emerged: assembly-line short dramas and low-cost productions used to fill platform content libraries are gradually being taken over by AI, with production speed and cost control becoming the core competitive metrics. By contrast, premium dramas centered on complex emotional expression and grounded realism are developing a distinct competitive advantage under pressure from AI. The unique atmosphere of a live set and the spontaneous emotional resonance that human performers bring are qualities that algorithms cannot replicate.

Xu Xiang, an AI film and television director and young filmmaker at the Shanghai Film Academy of Shanghai University, drew on his own experience to underscore the division of roles between humans and machines. He told CSST that “human-led, AI-empowered” collaboration is the core prerequisite for the healthy development of AI film and television. AI can generate images efficiently, but it cannot take part in genuinely creative decision-making. Script conception and storyboard design must remain under the control of screenwriters and directors. Creators must first be clear about what story they want to tell and why; AI’s role is to execute that vision efficiently at the visual level, handling auxiliary tasks such as scene construction and large-scale special-effects rendering.

Full-chain regulation in place

During their largely unregulated expansion, AI short dramas have accumulated a range of persistent problems. Unclear sourcing of training materials and infringements of portrait rights involving virtual characters have become increasingly common. Because infringements involving AI-generated content can be difficult to detect, while after-the-fact tracing and evidence collection are often extremely challenging, traditional copyright protection and content review systems have struggled to adapt to the technology’s distinctive production characteristics. These shortcomings have become a major obstacle to the industry’s healthy development.

Regulatory authorities are now moving quickly to close these institutional gaps and establish a clearer compliance framework for the regular prime-time broadcast of AI short dramas on major television channels. On July 31, the National Radio and Television Administration (NRTA) announced the “Administrative Measures for the Development of Micro Dramas,” which will officially take effect on September 1. The Measures explicitly bring AI-generated micro dramas under a unified regulatory framework, eliminating the previous regulatory blind spot that treated AI content as a special category. In areas including content review, intellectual property protection, and broadcast labeling, AI short dramas will now be subject to the same standardized regulation as live-action short dramas. A clear red line has thus been drawn.

Liu pointed out that, compared with live-action short dramas, the governance of AI short dramas presents a distinct set of challenges. Legal standards governing copyright and portrait rights remain ambiguous, making it difficult for existing rules to keep pace with AI’s batch-generation model. AI creation also raises new concerns involving aesthetic and privacy ethics. In the field of public literary and artistic communication, these governance issues are becoming increasingly urgent.

From a practical standpoint, Li proposed a compliance framework spanning the entire content production chain. At the upstream model-training stage, the use of copyrighted materials owned by the developer or materials in the public domain should be mandatory, while unauthorized scraping of film, television, and literary works for large-model training should be prohibited, preventing infringing materials from entering the process at the source.

At the scriptwriting stage, an originality verification mechanism should be established, with clear distinctions drawn between AI-assisted polishing and large-scale automated content generation. This would help reinforce the screenwriter’s primary role, while manual review could be used to prevent factual and logical hallucinations in AI-generated plots.

At the mid-production stage, the use of virtual portraits should be strictly regulated, with the replication of ordinary people’s or public figures’ facial features and voiceprints prohibited, while prompts and material-generation logs should be fully retained. At the submission and broadcast stage, satellite TV stations should establish dedicated review channels for AI content, making complete copyright traceability documentation a strict condition for submission, while also strengthening the screening of formulaic, template-based AI productions.