While studios and unions debate the long-term implications of artificial intelligence, early-career screenwriters in China are finding ways to incorporate the tools into their creative process, according to a recent study out of Hong Kong. The study from the University of Science and Technology found that most participants use tools like ChatGPT and Midjourney to brainstorm ideas, build characters and get creatively unstuck.
Presented this spring at the 2025 Conference on Human Factors in Computing Systems, the study surveyed 23 screenwriters, ages 23 to 29, in mainland China. Eighteen said they had used generative AI as part of their writing process — not to replace creativity, but to jump-start it.
The study appears to be the only academic research to date that closely examines how screenwriters are using AI. That it comes from researchers in China — not the U.S., where AI is being experimented with in areas like animation, VFX and marketing — may reflect Hollywood’s particular discomfort with applying the technology to screenwriting. Some fear generative AI applied to storytelling will only fuel mediocrity — and cut out human creators — in Hollywood.
In the Chinese study, writers described using AI tools to generate character backstories, sketch out plot points or visualize settings. One participant said they used Midjourney to generate an image of a terrified gorilla standing in a modern art museum. That unexpected combination inspired a speculative story about animals navigating human society. Others described similar moments of surprise — where the AI surfaced a detail or juxtaposition they wouldn’t have arrived at on their own.
“Most of them are positive, even though some will have concerns about the replacement of their jobs,” said Yuying Tang, the study’s lead author who researches how AI systems intersect with creative domains like film, design and fine art. “They also recognize that AI will likely improve with time.”
Tang added that participants viewed their experimentation with AI as a way to prepare for future changes in the creative landscape.
“They are trying to understand it early,” she said. “Sometimes they don’t directly use the output, but it gives them ideas.”
AI’s usefulness at this stage may lie in how it mimics the way human inspiration works — sparking loose associations that lead to new ideas, according to Zorana Ivcevic Pringle, a senior research scientist at the Yale Center for Emotional Intelligence. That’s why generative tools can help with brainstorming but fall short when it comes to more advanced tasks.
“Where AI is not useful is putting different ideas together and creating a coherent whole with nuanced understanding of emotions and social relationships,” she said. “We start with a coherent draft and iterate on it until it comes to life.”
The study supports that divide. While participants turned to AI to break creative blocks or generate raw material, few found the tools helpful for dialogue or structure. None of the six writers who tried generating dialogue said the results were usable. Of the 12 who used AI to shape narrative structure, only two said they would do it again. In some cases, the AI responses were described as stilted or formulaic — the kind of output that might pass for content, but not for craft.
Tang said that while the study focused on Chinese writers, the underlying creative challenges are similar elsewhere. “I don’t think there is a huge difference between different locations,” she said, “The needs are similar.”
Van Robichaux, a television and animation writer who served on the Writers Guild of America’s AI working group leading up to the 2023 strike, told TheWrap the study did a solid job capturing the structure of screenwriting, but overstated how widely generative tools are actually being used. He felt the authors leaned too heavily on academic citations and pointed to how the study framed the WGA strike as an example. “The concern was more where things were headed than where they were,” he said. “It wasn’t a reaction to widespread use.”
Still, he said the findings could hold relevance for Hollywood. The fundamentals of screenwriting — from development to revision — are largely consistent across borders. Rather than the location of the participants, Robichaux said their early-career status likely shaped their openness to AI.
“Younger people would be more willing to implement these things in their workflow,” he said. “But you quickly learn things work the way they work for a reason.”
He added that many of the ways participants used AI, such as for character generation or story prompts, don’t reflect how more experienced writers typically operate. Having worked on a project in China, he said he’d expect that distinction to hold there as well.
While Robichaux said he isn’t opposed to AI, he points out that the way these tools work — drawing from patterns in existing data — is fundamentally at odds with what writers are often trying to do: break formulas, not replicate them.
“One of the biggest concerns with AI-generated content is this averaging out,” he said. “It hits that ‘good enough’ bar but is pretty flat. We’ve already moved toward that on our own without augmentation.”
He sees the greater risk not in AI replacing writers, but in studios becoming comfortable with mediocre, machine-assisted work. “Hollywood already struggles with a lot of average storytelling,” he said. “We don’t need AI pushing us even further into that zone.”
