Abstract
The rapid growth of artificial intelligence is significantly transforming the production relationships and power dynamics within China’s creative and cultural industries. Focusing on small and medium-sized animation studios, this study uses participant observation and in-depth interviews to explore how AI technologies, within a platform-based environment, reshape the power balance between studios and platforms. The findings reveal that AI has shifted creative work from “drawing” to a practice mainly focused on correcting AI-generated outputs, effectively turning creators into “correctors.” Although efficiency has gone up, work has become more demanding, and the risk of losing skills has increased. This change results from the technological monopoly platforms built with proprietary AI tools, shifting small and medium-sized teams from traditional “distribution dependence” to a dual reliance on “distribution channels and production tools.” Under this new monopoly, workers’ operational data is used as training material without compensation, leading to implicit exploitation and the erosion of studios’ core knowledge assets, which weakens their unique competitive advantages. In response, small and medium-sized teams develop strategies such as “manual refinement,” creating a continuous bargaining dynamic; however, these efforts are ultimately absorbed by platforms as training data. This study reveals a new type of platform monopoly in the AI era, one that extends beyond market dominance to also control creative practices and data extraction.
Keywords
Introduction
Artificial intelligence (AI) has rapidly entered animation production and is widely seen as a means to reduce creative barriers and enhance efficiency. While AI’s application to modeling, rendering, and content generation is often cited as automating repetitive tasks and improving production speed (Anantrasirichai & Bull, 2022), closer scrutiny reveals a complex dynamic. Industry and policy narratives frame this as “technological empowerment,” suggesting intelligent tools may equalize technical capabilities among animation firms (Jahromi & Ghazinoory, 2024). Yet within China’s highly platformized cultural industry, technological integration has reinforced rather than decentralized power. Deep AI integration into platforms enables them to increasingly direct content production via algorithms, data leverage, and distribution infrastructures, thereby restructuring creative industry dynamics (Gillespie, 2018; Poell et al., 2021).
The animation industry is a crucial field for observing this phenomenon. It has among the highest digitization and process orientation in the cultural and creative industries. Animation production includes technology-intensive stages such as modeling, motion capture, rendering, and visual generation. These provide abundant data and practical opportunities for training machine learning models and applying AI (Anantrasirichai & Bull, 2022; Feng & Yang, 2024). Meanwhile, the cultural and creative industries operate in value networks of multiple actors. Small and medium creative teams often drive original ideas and artistic experimentation, serving as nodes that sustain innovation and aesthetic diversity.
As AI technology enters early creative processes and style generation, its impact goes beyond improving production efficiency. It directly affects the positions of creative actors within the division of labor and the distribution of value. Therefore, understanding the ecological changes and response strategies of small and medium-sized creative teams under AI intervention is crucial. These insights help us see how AI influences the industrial ecology and power relations in China’s animation industry. This study focuses on the evolving dynamics and challenges of these teams as they adopt AI.
This study addresses three questions: (1) What impact has artificial intelligence had on creator practices in China’s animation industry? (2) Does AI help small teams overcome persistent or emerging structural difficulties? (3) How do small and medium creators survive in the new ecology? To explore these questions, this paper conceptualizes the transformation of creators’ practices as a shift from creators to “correctors,” where work increasingly focuses on the continuous selection, modification, and refinement of AI-generated outputs.
Based on 10 months of participant observation with the Youku AI storyboard research and development team and on semi-structured interviews with 25 animation practitioners, this study finds several key trends. While AI tools enhance production efficiency and accelerate the realization of creative ideas for small and medium-sized teams, platforms embed their self-developed AI tools into the production process or require their use. This integration makes the labor of these teams less visible. It also extracts more cognitive labor from workers and appropriates more value from their outputs. This reinforces the platforms’ dominant power positions.
The empirical findings are organized into three core sections, each addressing one of the study’s primary research questions. By demonstrating how AI tools intervene in the animation industry, the analysis gradually reveals the power struggles between platforms and small- and medium-sized creative teams within evolving industrial ecosystems.
The first section begins with a workflow perspective, examining how platforms integrate AI tools into the production process, thereby transforming creators’ practices and the working models of small- and medium-sized teams, ultimately impacting the overall ecology of China’s animation industry. The second section takes a creator-practices perspective, analyzing the challenges faced by small- and medium-sized teams, the new industrial environment shaped by AI, and whether their situation has improved or worsened in this context. The third section aims to identify the survival and bargaining strategies that small- and medium-sized teams use in this new environment to overcome the difficulties they encounter.
China’s Animation Industry and Artificial Intelligence
From 1949 to the early 1980s, animation production in China was primarily organized under a state-led studio system, with state-owned film studios handling production. The most representative among them was the Shanghai Animation Film Studio. Under this system, animation was typically created through collective efforts within the studio. The main goals of the production were cultural dissemination, artistic exploration, and educational purposes, rather than focusing on commercial markets (Lent, 2001).
After the Reform and Opening-Up in 1978, especially during the mid-to-late 1980s, China gradually became part of the global animation production network (Lent, 2001). In 1985, Jade Animation, backed by Hong Kong’s TVB, was established in Shenzhen and became the first professional animation company on the mainland. Later, processing companies like Pacific Animation, Cailing Animation, Anli Animation, Zhuhai Chisheng Cartoon, and Jing Hongye Animation were set up in the Pearl River Delta, drawing in animation talent from across the country (Niu, 2011).
In 1986, Taiwan-funded Hongying Animation was established and later moved to mainland China in 1992, where it founded Suzhou Hongying. In 1988, the first private animation company in mainland China, Dalian Avanti International Animation, was created, initially doing outsourcing work for Europe and Japan (Niu, 2011).
By the 1990s, the industrial landscape had expanded to two main regions: the Pearl River Delta and the Yangtze River Delta. Along the southeastern coast, there were 5–6 large animation companies, loosely spread out, and over 80 smaller ones, with the two largest being Taiwan-funded companies Wang Film Productions and Hongying. These businesses mainly focused on labor-intensive tasks like in-between drawing and coloring, forming the basic industry structure where large international companies led, while small domestic studios were scattered throughout the industry (The evolution of Chinese animation outsourcing, 2013).
China subsequently replaced South Korea and Taiwan as the world’s leading animation outsourcing hub. Over 70% of Japanese animation and many European and American animated productions completed their intermediate stages in China. Jade Animation contributed to Hollywood productions such as The Lion King and Mulan (Niu, 2011).
In 2004, the animation industry was officially included in the national cultural industry development strategy agenda. Animation output then increased rapidly, rising from about 12,000 minutes in 2003 to approximately 220,000 minutes in 2010, and in 2006 it surpassed Japan to become the country with the highest animation production output in the world (Su, 2022).
Since the mid-2010s, with the rise of digital platforms, China’s animation industry has undergone another structural shift. Online video platforms have increasingly become essential infrastructures for producing cultural content. By investing in animation projects, developing IP, and integrating distribution channels, they have significantly shaped the content creation process and continuously strengthened their role within the industry value chain (Cunningham & Craig, 2019; Keane, 2013; Oreate, 2026). Industry reports show that leading platforms have become the main distribution channels for domestic animation, with iQIYI, Tencent Video, and Youku collectively accounting for 87.8% of the market share (Vzkoo, 2023).
Within this platform-driven production network, although several large animation companies—such as Xuanji Technology, Rocen Digital, Haoliners Animation, YHKT Entertainment, and Phantom Digital—have emerged and work with platforms or IP rights holders on a project basis to produce content, it is the numerous small and medium-sized production teams that actually handle most domestic animation tasks. These teams usually subcontract for larger companies or platforms, making them a vital force within the platform-led model.
This division of labor, where “platforms control the channels while small teams handle production,” combined with the animation industry’s long-term reliance on many labor-intensive, repetitive tasks, makes it one of the sectors most affected by artificial intelligence (Barber et al., 2016; Torrejon et al., 2020). With advances in technologies such as machine learning and deep learning, reports indicate that AI platforms now manage the entire process from concept design to 3D rendering, including sketch generation, intelligent in-betweening, and automatic rendering (Anantrasirichai & Bull, 2022; IvyPanda, 2018; Tang et al., 2025). This also shows that AI technology is systematically transforming the animation industry. Industry reports show that about 70% of small and medium-sized animation studios have already adopted AI tools for asset creation, and the adoption of AI animation tools has cut the average production cycle by roughly 30% (SuperAGI Research, 2025; WiFiTalents, 2025).
This technological shift has created new opportunities for small- and medium-sized animation teams in China, which often struggle with limited resources. A clear example is the Chinese AIGC-based animated series “Tomorrow Is Monday.” This project was completed by a team of just 10 people in 45 days and produced 50 episodes, with the cost per minute dropping significantly from the traditional 30,000–50,000 RMB to only 2,000–3,000 RMB. (Sina Finance, 2025).
Platformization, Technological Monopoly and Artificial Intelligence
Platformization theory examines how digital platforms influence and reshape content creation and value sharing through technological structures, algorithmic systems, and rule frameworks (Helmond, 2015). Platforms are not neutral tools but governance systems with defined power dynamics. They first bring various participants into established rules via programmable architectures, using this as a basis for expanding capital and extracting data. Then, they serve as “curators” through algorithmic systems in content distribution and visibility, further shaping the norms of cultural production (Gillespie, 2018).
Although some studies highlight the positive role of platforms in resource sharing and innovation diffusion (Chen et al., 2022; Dahmani & Ben Youssef, 2023; Jiang & Sun, 2025), critical perspectives point out that such sharing often depends on platforms monopolizing data profits. This unequal power dynamic has been well documented in the distribution aspects of creative industries. Research on Spotify shows that algorithmic recommendation systems strip creators of control over their work’s exposure, forcing them to passively adapt to algorithmic rules (Prey, 2020). Similarly, research on the Webtoon platform indicates that algorithmic ranking and crowdsourcing mechanisms enable platforms to define value, resulting in a high concentration of revenue among top creators, while most creators become marginalized and passive recipients of platform rules (Kim & Yu, 2019).
In modern research, AI is widely seen as the next phase of platformization. By controlling essential resources like data, models, and computing power, AI has significantly changed how creative work is organized and how value is distributed (Berlinski et al., 2024; Fox, 2024). Optimistic studies emphasize how AI transforms the relationship between platforms and creators, arguing that AI tools lower the barriers to creation and enable more small teams and individual creators to join platforms and participate in content production, opening new opportunities for creators (Capraro et al., 2024). However, some research also highlights that this kind of technological monopoly can actually strengthen platform control over infrastructure and shift platforms from relatively open innovation systems to highly centralized power structures (Verdegem, 2024).
In this paper, “technological monopoly” refers to platform-based control over production tools, data infrastructures, and algorithmic systems, through which platforms extract, standardize, and reappropriate creators’ practices as data for model training. The idea of technological monopoly goes beyond market dominance in economics and is more deeply expressed as cultural and ideological control (Illich & Lang, 1973; Postman, 2011). Langdon Winner (2020) noted that technological artifacts are not neutral; specific technological designs reflect particular power preferences. Neil Postman (2011) described this as the point where a technology becomes dominant in an industry, shifting from an optional device to the only standard for defining “progress,” “efficiency,” and even “truth.”
In the digital era, the most common practitioners of technological monopoly are digital platforms that combine computing power, data, and distribution channels (UK Competition and Markets Authority, 2015; Zuboff, 2023). For creative industries, leading platforms create insurmountable knowledge barriers by controlling underlying algorithms, training data, and computing resources, forming what is known as an “Obligatory Passage Point” (Verdegem, 2024), making the platform the essential technological gateway through which creative work must pass. Nick Srnicek (2017) observed that platforms establish a new monopoly over technological infrastructure by controlling data and user connections. Zuboff (2019) further described this as a “Radical Monopoly,” where small actors lacking such technological capabilities are deprived of legitimacy in the market. Nieborg and Poell (2018) uncovered the concrete operation of this power structure from the perspective of creators: platforms monopolize content circulation by controlling algorithms and distribution tools, and creators must follow platform rules to gain visibility. This exclusive control over access to audiences raises technology from a mere tool to a form of power, locking creators into a dependency structure.
Research Methods
This study uses participant observation and semi-structured individual interviews. In research on creative and cultural industries, Hesmondhalgh and Baker (2013) highlight the importance of qualitative methods that combine participant observation and semi-structured interviews to gain a deeper understanding of everyday practices in creative work. Participant observation allows researchers to access production sites and gather first-hand information about workflows and interaction patterns, capturing tacit knowledge and behavioral details that are hard to uncover through textual or statistical data (Musante & DeWalt, 2010). Semi-structured interviews give researchers a way to understand actors’ intentions, helping them explore practitioners’ interpretations and experiences of their own practices. These two methods complement each other: observation traces behavior, while interviews reveal its underlying logic. By combining these approaches, this study achieves methodological complementarity and cross-validation.
This study was conducted collaboratively by multiple authors. First, the second author (also the corresponding author) conducted a 10-month period of participant observation within the Youku team, a leading platform under Alibaba, from September 2024 to June 2025. Supported by Alibaba’s capital, the Youku team is not only one of China’s top video distribution platforms but also a comprehensive platform that includes multiple departments such as animation production, technology research and development, and algorithm construction, with as many as 65 contracted animation teams. Compared to companies that rely on third-party tools or only participate in content production, the practical experience of employees within the platform better demonstrates how the platform influences the production process through technological tools, reorganizes the division of labor, and accumulates data resources.
The second author gained access to participant observation through an internship with the Youku team and conducted the research with the team leader’s knowledge and permission (see the “Ethics Statement” section below for details on ethical approval and procedures). During the research, the second author primarily engaged in participant observation within the natural work environment of the Youku team. They strategically positioned themselves in various roles across different stages of fieldwork, drawing on their current role as a researcher and their previous experience as an animation planner and participant in AI animation production projects. The company was informed about the nature of the research and consented to the use of relevant materials under conditions of anonymity and the exclusion of sensitive information. Ultimately, because of the team members’ understanding of the second author’s researcher identity and prior professional experience, along with the trust built through long-term interaction, the second author was able to participate in data testing and content review within the AI storyboard R&D group, enabling in-depth observation of the development and integration of platform AI tools.1,2
Based on participant observation, the second and third authors also conducted semi-structured interviews. The industry connections made during the observation provided important access points for the follow-up interviews. The interviewees were divided into two groups. The first group included members from different departments and with varying levels of experience within the Youku team, covering roles such as content production, technological research and development, project management, and platform operations, totaling 11 people. The second group consisted of members from small and medium-sized animation studios that have partnerships with the platform, contacted through industry connections established during fieldwork. These included animation directors, storyboard artists, key animators, and project producers, totaling 14 individuals. All interviewees received an info sheet before the interviews and gave clear consent for their responses to be used for academic research and publication. All data were anonymized (see the “Ethics Statement” section below for details).
These studios are usually situated in the middle and lower parts of the platform-led animation industry value chain and handle specific production tasks like storyboard creation, key animation, 3D modeling, and post-production. They thus gain direct experience using platform technological tools, project management techniques, and changes in labor division. This study focused on selecting production teams that have cooperative or outsourcing ties with platforms as interview subjects, including studios like Xuanji Technology, which has participated in well-known domestic animation projects such as The Legend of Luo Xiaohei. These teams are both representative within the industry and located at key points in the platformized production network. Their practitioners help illustrate the cooperation patterns between platforms and production teams, as well as how platformization influences specific production processes.
The interviews were conducted using purposive sampling (Goodman, 1961; Palinkas et al., 2015; Patton, 2014). A total of 25 semi-structured interviews were conducted, each lasting between 45 and 90 minutes. These interviews took place in both face-to-face and online formats. The interview guide centered on the following key themes: (1) The collaboration models between platforms and production studios; (2) The ways in which AI tools are integrated into animation production workflows; (3) The influence of AI on the division of labor and creative autonomy. All interviews were audio-recorded and transcribed into text. During the interviews, the researchers emphasized the necessity of using clear and precise language, avoiding ambiguous or vague expressions to ensure communication accuracy. The data were then coded and analyzed thematically by the research team. For Chinese-language interview materials, relevant excerpts were translated into English by one author based on the analysis, and then reviewed multiple times by several authors to maintain the original context and semantic nuances as much as possible.
Research Discussion
The Ecological Impact of AI on the Animation Industry
The current animation industry in China displays a typical “pyramid” structure. At the top are leading platform companies such as Tencent Animation and Bilibili, which control key distribution channels and IP resources and hold a dominant market position. In the middle are large animation production companies like Xuanji Technology, Original Force Animation, Light Chaser Animation, and Fantawild, which have full-process production capabilities and consistent output. At the base are numerous small- and medium-sized studios—estimated to exceed 6,000—most of which depend on outsourced projects to survive (Liu, 2014, 2025; Xiong, 2025).
Unlike the first two tiers, small and medium-sized studios usually do not engage in core IP development or full-process production. Instead, they focus on modular tasks like scene asset creation, visual effects rendering, and in-between frame drawing, leveraging their specialized skills. They embed themselves into platform-led production networks on a project basis, acting as the “capillaries” of the industry chain (Li, 2011). In terms of collaboration, platforms connect large production companies and small studios through equity investment and project-based cooperation (Li, 2011; Liu, 2014). These small studios handle most labor-intensive tasks, such as rendering and in-betweening. For example, in Ne Zha 2, among 138 participating companies, small- and medium-sized teams accounted for over 87% and completed most of the film’s production workload (Wu & Wu, 2025).
This division of labor, in which small and medium-sized studios serve as the main workers, has led to many repetitive, standardized, and labor-intensive tasks persisting in animation production. These tasks are now the first to be affected and changed by artificial intelligence. AI is now entering China’s animation industry at an unprecedented pace, with the changes at the workflow level being immediate and highly noticeable.
Repetitive and detailed tasks that once needed a lot of human effort can now be done efficiently by algorithms, and features like “text-to-storyboard” are changing the early steps of animation production. However, behind what technological optimists call “new opportunities” for entering the industry (Capraro et al., 2024), a more complicated reality is starting to show: the heart of creative work—the artistic intuition built up over many years, the aesthetic judgment honed through practice, and the unique style developed through repeated refinement—is being reshaped by technology.
First, on a technical level, by streamlining workflows, AI tools help small and medium-sized teams boost production efficiency and lower the costs associated with trial-and-error from early-stage design through later-stage rendering. For creators, this efficiency gain is more noticeable as a shift in labor. The studio where Oliver works has already achieved full-process AI-assisted animation generation by building workflows on an AI platform, and her primary work has shifted from “drawing” to “image correction.” She explained: “Basically, there is no need to draw. The main issue is that AI-generated content contains errors, such as four fingers or similar mistakes. We need to review them and modify the prompts to correct the mistakes.”
It is worth noting that the “relief” AI has provided has not truly freed human workers. The interviewed animators and storyboard artists both stated: “The workload has become smaller, but the work intensity has in fact intensified.” The reason is that although the process from creative idea to completed work has been shortened, labor intensity has not decreased accordingly. As the animator Alice remarked, “Before (as an animator), we drew about 3 to 4 images per week. In the studio, each person must produce at least 5 images per day. Hand drawing is no longer required, but when you use AI, the company’s expectations for your output become much higher.” It can therefore be argued that the time saved by technology has not been turned into more personal time for workers; instead, it has been redirected to meet higher productivity demands.
Another animation storyboard artist, Jimmy, added, “When we were hand-drawing before, the process was actually more flexible. Once you finished drawing, you were done. Now, every day you’re tied to the workstation. You’re either writing prompts or repeatedly modifying prompts based on the images produced by AI—adjust the prompt, generate an image, it doesn’t work, change another set of keywords, generate it again. AI can produce an image in a few seconds, but getting a usable one may take dozens of iterations. On top of that, bosses think AI is efficient, that one person can replace several people, so the number of assigned tasks keeps doubling. There’s scarcely any breathing space.”
This feeling isn’t unique to Jimmy. For many creative workers, this process increasingly positions them as correctors, whose work involves continuous prompt adjustments, error corrections, and iterative refinement. Increasingly, people are experiencing the overwhelming sensation of being “dragged along by machines and unable to stop even for a moment.” On the surface, technological progress has made image generation faster, but in reality, the workload is quietly growing—constantly writing prompts, fixing errors repeatedly, and making continuous revisions—all of which become additional burdens that individuals must silently carry.
Even more hidden is that in the process of repeatedly correcting AI errors, creators contribute not only physical effort and aesthetic judgment but also provide training data for AI models without compensation, using their own labor to help machines become “smarter,” while they themselves fall into an even more intensive cycle of work. These changes in the individual work content of laborers are not merely isolated shifts in professional experience but signal a fundamental restructuring of the production model of small and medium-sized teams. The former “workshop-style” production, which relied on internal knowledge transmission and accumulated experience within teams, is gradually being replaced by “standardized interface-style” production defined by platform AI tools. Core team members shift from “creators” to “correctors,” while the production rhythm, output standards, and even the team’s creative direction are shaped by the technological frameworks set by platform tools.
Changes in workflow have also directly influenced the industry’s occupational structure, with entry-level and repetitive roles among the first to be affected. In animation production, AI technology has predominantly integrated into tasks that are relatively standardized and easy to automate, such as bulk generation of scene assets or assistance with in-between frame creation, demonstrating a clear focus on cost efficiency. This trend was supported by several interviewees: Eleanor mentioned that the number of employees in the same role has decreased, while senior lead animator Oliver noted that the impact has been concentrated on “entry-level positions.” Likewise, Samuel, who works in design, reported that the promotional poster department has been reduced from 10 people to two, plus AI.
Contrary to the common expectation that many “prompt engineers” would emerge, new types of roles have not appeared. Instead, people in their current jobs need to learn new skills to stay up to date. For example, John, the technical manager of Alibaba Youku’s AI storyboard team, said, “In the past, when we hired people, we mainly looked at professional ability. But now it’s different—we prefer people who not only have a solid professional background but can also skillfully use AI tools.”
The impact of AI on the animation industry has two sides. On one side, it boosts production efficiency and cuts costs. On the other side, it quietly reshapes the core of creative work—artistic intuition built up by creators over many years, aesthetic judgment refined through practice, and drawing skills internalized through repeated effort are being redefined by technology in every act of “correction” and “selection” during daily work. These two changes happen at the same time and together form the structural context faced by small and medium-sized teams in the age of AI.
Constraints of Small and Medium-Sized Teams: Mitigation or Intensification?
Before artificial intelligence transformed the animation production process, small and medium-sized animation teams in China had long faced structural challenges in their path toward industrialization. Due to limited control over upstream core IP and downstream distribution channels, these teams have historically been pushed into the low-profit middle segment of the industry chain, mainly handling outsourced or commissioned work, with their survival space and bargaining power continuously constrained (Pan, 2024; Toubao Research Institute, 2020).
With the rise of streaming platforms, platform capital has further strengthened its dominance over production by controlling distribution channels and traffic allocation. Although small teams have gained more opportunities for exposure, they are also compelled to adapt to the platforms’ algorithmic logic and traffic rules. Consequently, their creative direction increasingly compromises with market demands, while the gap between value contribution and economic return continues to widen (Xu, 2024).
Under the influence of AI, a notable trend is that while small and medium-sized teams use platform AI tools to boost production efficiency, they may also face the risk of gradually losing their core competitiveness. In traditional animation production, small and medium-sized teams build experience, technology, and aesthetic preferences through project practice, transforming these into reusable internal knowledge assets that are crucial for their competitiveness. When AI simplifies work steps, the project execution process bypasses the repeated refinements that were once necessary, depriving team members of opportunities to gain experience and develop personal aesthetics through continuous revision.
A senior designer deeply felt this change: “In the past, most of the day was spent struggling with design drafts and prototypes; now it might take only 2 hours to get it done.” In his view, this is not because people have become lazier, but because the rhythm of production has been completely changed by AI. This also means that when the process of repeated refinement is removed, creators lose the opportunity to accumulate experience and develop personal aesthetics through continuous modification.
The transformation of work steps has also led to the devaluation of creators’ labor. Betty, a screenwriter, wondered, “Now my work basically involves writing prompts based on tasks and filtering the results. This makes me think: what is my unique value in this creative process?” Betty’s confusion is not an isolated case. When creative work is reduced to two stages—“writing prompts” and “filtering results”—the artistic intuition built up by creators over many years is displaced. Related research indicates that artificial intelligence capitalism reinforces the logic of efficiency and scale, and may alter the system of value assessment in cultural production (Verdegem, 2024). Professional aesthetic judgment shifts from that of a creator to that of an inspector, thereby reducing the value attributed to creative contribution in the production process. This devaluation occurs because AI has reshaped the foundations of value in animation, moving the focus from originality and artistry to adaptability and efficiency.
Based on the resource-based view (Barney, 1991), the real competitive advantage of small teams should be creative outputs that are “difficult to imitate.” In the traditional model, this advantage was built by teams gradually gaining experience, technology, and aesthetic preferences through long-term practice, turning them into internal knowledge assets. However, AI as a tool is dissolving this mechanism.
In the traditional model, production tools were fairly neutral. Over time, teams could gradually develop their own technical habits and aesthetic preferences. However, today’s AI tools are different: they are mostly standardized services offered by platforms, and everyone uses the same models and similar parameters. When tools no longer allow teams to incorporate their own experience into the creative process, the advantages they once relied on become harder to leverage, and maintaining uniqueness becomes more difficult.
Several interviewees confirmed this trend. Robert pointed out that AI tools generally tend to produce “template-style aesthetics.” Paul, an independent animator with many years of experience, reflected: “Bosses are eager to use AI to cut costs and boost efficiency. The work is indeed done faster, but the outputs may eventually look the same.” The 29-year-old animator Judy described her daily work like this: “Hand drawing used to be the skill I relied on to make a living. How the lines move and how the shapes are captured were all in my own hands. Now I no longer need to draw by hand—animation generation relies entirely on Stable Diffusion.”
Judy’s observation is not just about replacing tools. Instead, optimizing production tools and platforms encourages creators to abandon their long-held skills and rely heavily on platform-specific tools. While this may seem empowering, it results in a form of hidden technological dependence. As creators increasingly rely on platform infrastructures for both production and distribution, cultural creation becomes gradually integrated into the system of platformization (Nieborg & Poell, 2018). In such a setup, when technological tools become the primary means of solving problems, creators’ reliance on a particular technological system can evolve into a technological monopoly (Illich & Lang, 1973).
Judy’s personal experience highlights a broader collective problem: when team members stop developing new skills, the organization’s core knowledge assets lose their ability to regenerate. From a resource-based view, the real competitive edge for small and medium-sized teams should come from “tacit knowledge” embedded in members’ practices, which is hard for the market to copy. But AI tools standardize and clarify the creative process, turning experiences that were once hidden within teams into explicit data that platform algorithms can capture. What the team loses is not just individual skills but also the organization’s ability to learn and its unique knowledge reserves.
A manager of a medium-sized animation studio, Kim, said: “We do feel that something is a bit strange, but we can only use their tools. We basically have no say in it. After all, the platform provides them for us to use for free, and we don’t really have any position to refuse.” From Kim’s words, we see that promoting platform AI tools is not just about creators relying on them; it’s about deliberately deploying tools at the production end. Creators are aware of this, but because of their disadvantaged position, they have no direct way to oppose the platform’s embedding of these tools.
By monopolizing production tools, platforms absorb creators’ aesthetic knowledge and work as AI training data, transforming their labor into the core resource that drives model iteration. Creators contribute cognition and creativity but do not receive a proportional return. Meanwhile, platforms continue collecting data and expanding their control over production, further solidifying their monopolistic dominance.
A manager of a medium-sized animation studio stated, “We often cooperate with the platform to test some of the AI tools they develop. We give them a lot of feedback, which is also considered part of our work, but there is no payment.” In the Youku AI storyboard project observed in this study, project managers frequently invited small- and medium-sized studios to test AI tools for free and used their feedback to guide subsequent modifications. Additionally, part of the model’s training data comes from the internet, while another part comes from works created by small and medium-sized teams, and the project team has never paid copyright fees to either party.
The data engineer, Robert, explained, “Other teams (other AI tool development teams) are doing the same thing. Since this is for storyboarding, the outputs generated by AI will not retain traces of the original content, so we are not considering copyright issues at the moment.” Another storyboard artist, Amy, added, “We have never considered income from this aspect. Since we cooperate with the platform, it would not be good to fall out over such a small matter. We just treat it as something done along the way.”
When faced with the platform’s data extraction, small and medium-sized studios and creators experience a clear compromise and helplessness. Creators are personally training and refining the AI tools that might replace them someday or further diminish their bargaining power. By monopolizing production tools, platforms absorb creators’ surplus value into their models, making the platforms increasingly dominant while the individual value of creators continues to be diluted.
The Survival Strategies of Small and Medium-Sized Creators Amid Emerging Trends
The introduction of AI technology has greatly increased the complexity of the bargaining relationship between small teams and platforms. Previously, platforms controlled traffic distribution, and small teams depended on them to reach audiences, mainly negotiating over IP ownership and revenue sharing (Nieborg & Poell, 2018; Poell et al., 2021). AI has changed this dynamic, shifting small teams from a simple reliance on distribution channels to a dual dependence on both “channels and production tools.” This shift makes the situation more delicate for small teams—they must rely on platform tools to boost efficiency while also actively minimizing the “AI traces” in their work, aiming to showcase their creative value by maintaining a sense of craftsmanship and emphasizing personal style.
To meet platform requirements for AI tools, most small and medium-sized teams interviewed adopt a production strategy that combines “manual refinement” and “AI generation,” with the manual-to-AI ratio roughly 4:6. Using the team where animator Peter works as an example, he explained: “Even when clients require the use of AI, we still try in actual production to ‘minimize visible AI traces as much as possible’.” For these teams, this isn’t just an artistic stance but also a survival strategy in the market. By emphasizing manual refinement and the personal style of contracted animators, they aim to distinguish between “pure AI generation” and “partial AI participation” in the marketplace. In doing so, they try to establish a different position, marked by a “sense of craftsmanship,” beyond the platform logic that prioritizes efficiency above all else.
The reason teams dedicate significant effort to “manual refinement” of fine details is to show the platform that high-quality results come not from the machine’s automatic output but from the team’s judgment and craftsmanship—elements that algorithms cannot replace. This careful manual refinement, across many details, becomes a bargaining chip that demonstrates the team’s value to the platform. Labor process theory highlights that under “deskilling” pressure, workers attempt to regain control through their labor traces, which also serve as resources for showcasing their value to capital (Burawoy, 2012; Edwards, 1982). At the team level, this principle is reflected in how teams turn the individual skills and labor traces of contracted animators into differentiated competitive advantages at the organizational level.
It is crucial to understand that the bargaining relationship between small- and medium-sized teams and platforms exists within a larger dynamic. On one side, teams depend on their animators to resist standardization through “manual refinement,” aiming to showcase the team’s unique value through independent creation. On the other side, platforms do not suppress this autonomy but instead turn it into a key resource for production—every instance of “manual refinement” by animators, including operational traces, modification records, and creative decisions, becomes training data for platform algorithms. This means that the more teams rely on animators to demonstrate their irreplaceability through “manual refinement,” the more they provide platforms with examples of how these skills can be duplicated or replaced. The very tool teams use to combat standardization, while serving as leverage, also accelerates the digital breakdown and reproduction of the craft itself.
Animation planner Lindy offered supporting evidence: “Now, some clients do actively request that the work ‘retain a certain hand-drawn feeling’ or ‘not have too strong an AI flavor’. They worry that work generated entirely by AI may appear cheap.” When platforms produce promotional materials, they often require production teams to submit process-oriented materials, such as hand-drawn sketches and modeling assets. These are included in the publicity content. A liaison from one platform explained: “It’s more about showcasing a kind of craftsmanship, but it can’t be used as an excuse for delaying deadlines. Our requirement is always to ensure product quality while meeting delivery schedules.” Capital thus demands that creators maintain efficiency while simultaneously conveying a sense of craft. This dual pressure—of “efficiency” and “human touch”—amounts to a form of double extraction. Animators, under severely compressed timelines, must labor intensively to erase traces of the machine.
Hesmondhalgh and Baker (2013) argue that creative work typically offers a high level of autonomy, enabling workers to express themselves, experience a sense of internal achievement, and gain recognition from peers. When faced with a loss of a sense of achievement due to skill deterioration that is hard to reverse in their work, creative workers shift their creative focus to their leisure time. This phenomenon, called leisure crafting, acts as a psychological compensation and self-protection mechanism that creative workers instinctively develop when their work becomes “deskilling.” To some degree, this mechanism fulfills creative cultural workers’ need for achievement and helps prevent the decline of their core original skills (Petrou & Bakker, 2016; Petrou et al., 2024). Several interviewed animators stated: “Daytime is work, after work is life. When I go home, I draw things I like, and it actually gives me a sense of control over life.” Since high-intensity, low-skilled routine labor during the day no longer provides a sense of achievement or aesthetic satisfaction, creative workers actively pursue highly autonomous, challenging creative activities after hours to counteract the feelings of meaninglessness from their daytime jobs. One animator said, “I definitely can’t spend my whole life being this kind of technical laborer. Drawing something after work is also a way of exploring the direction of my next job, to see whether there might be other opportunities.” Another animator believed: “AI animation is too low-end and has little value; it will definitely be eliminated. After work, I still keep drawing to maintain my skills. Hand-drawn animation will sooner or later become mainstream again.”
Unlike established animation teams, some newly formed teams try to adopt AI with optimism in the early stages of their creative process. For them, AI is a way to boost revenue and cut costs. However, without support from platform capital, these startup teams often face high subscription fees for AI tools. One interviewed animation director mentioned that they use strategies like “sharing accounts” or “seeking alternative AI tools,” and control costs by “making the most of each use and choosing only the most valuable parts.”
In our research, we also found that as barriers to using AI tools decrease, many people with no previous experience have started experimenting with AI to create animation. One hobbyist interviewee said: “Now that AI is so advanced, I create AI animation partly because I’m worried about losing my job in middle age and want to learn more technical skills, and partly out of personal interest.” However, these individual AI animation accounts often update slowly. Several interviewees stopped updating entirely after two to 3 months, and some respondents mentioned that “it’s not interesting and there’s no way to make money” or “work schedules do not allow it” as the main reasons for giving up.
In summary, small and medium-sized teams adopt dual survival strategies. They rely on AI tools from platforms to maintain production efficiency while also trying to demonstrate their unique value by emphasizing manual work. However, all of this occurs within the rules established by platforms. The more teams showcase their individuality, the more they provide training samples that help platforms standardize and automate these traits. Therefore, efforts by teams to resist and by platforms to absorb happen simultaneously in the reality of AI for small- and medium-sized creative actors.
Conclusion
Through participant observation and in-depth interviews with small- and medium-sized animation studios in China, this study reveals that the spread of artificial intelligence technology in China’s animation industry is a dynamic process characterized by tensions. It is neither purely technological empowerment nor a one-sided technological monopoly. Instead, it is an ongoing practice in which platform capital, through technological tools and its interactions with small- and medium-sized creators, continuously reshapes production relationships and value distribution patterns within the industry. The study finds that the diffusion of AI technology in China’s animation industry follows a clear path: initially viewed by small teams as a way to improve efficiency and cut costs, it gradually develops into a technological monopoly resource through which platforms strengthen their dominance.
This study demonstrates that AI is transforming China’s animation industry at the workflow level. While enhancing production efficiency and shortening production cycles, AI has also significantly altered the types of creator practices within the industry. Repetitive and standardized entry-level roles are being greatly reduced—tasks like scene creation and in-between frame completion are now handled by algorithms, and in some teams, the number of workers performing similar roles has been cut by more than half. Meanwhile, the focus of positions such as animators and storyboard artists has shifted from “drawing by hand” to “AI generation with manual correction.” They no longer draw frame by frame; instead, they spend more time refining prompts, comparing outputs, and performing similar tasks. Technology has not truly reduced labor intensity; rather, higher production targets have increased the workload for remaining workers.
More importantly, the benefits of this efficiency are not shared equally. By integrating their self-developed AI tools into the production process, platforms establish new technological rules and production rhythms. As a result, while small teams experience efficiency gains, they are also compelled to adopt the technological architectures defined by these platforms. The non-neutral nature of this technological architecture becomes a central issue in the restructuring of power relations. This means that what AI technology reshapes is not only the working methods of individual laborers but also the foundation for the survival of small and medium-sized teams as creative production organizations: teams can no longer differentiate themselves through the accumulation of internal knowledge; their production processes are deeply influenced by platform tools, and their core assets are continually transformed into model resources for the platform.
Regarding the structural challenges faced by small teams, this study finds that AI intervention creates a dialectical effect where relief and intensification coexist. On the one hand, AI tools reduce technical barriers, allowing resource-limited small teams to complete production tasks that were previously difficult for them to handle on their own. On the other side, this convenience comes at the expense of losing core competitiveness. Relying on standardized AI tools provided by platforms strips teams of the opportunity to develop internal knowledge assets, leading not only to the blending of artistic styles but also to further diminishing their bargaining power.
Furthermore, platforms absorb the creative outputs of small teams as training data for AI, creating a form of “implicit exploitation.” Creators are enhancing technological tools that could eventually replace them, yet they are not compensated for this contribution. In this process, the value of their labor is extracted a second time, while platforms continue to strengthen their technological monopoly through ongoing data collection.
Additionally, regarding the strategies employed by small creators in the new ecosystem, the study finds that they do not passively accept technological change but show active agency within limited conditions. When faced with platform demands to use AI tools, creators try to maintain a sense of craftsmanship and strengthen their personal style through “manual refinement,” aiming to demonstrate their irreplaceable creative value within a system that values efficiency above all. This act of resistance is both a response to the pressure of “deskilling” imposed by capital and a small-scale effort to contest control over the labor process. Creators attempt to keep their unique relevance by investing in custom, non-standardized work. However, these differentiated outputs ultimately become essential data for the ongoing development of AI systems. This creates a paradox: the more creators seek unique expression, the more their work is used as samples that improve the algorithm’s ability to imitate. In this cycle, the personal experience that once set creators apart from machines is absorbed by the system and turned into resources that boost automation, gradually making creators structurally serve as input sources for AI platform systems. Meanwhile, feeling anxious about skill loss from mechanized work during the day, some creators pursue highly autonomous leisure crafting after hours, trying to counteract their skill decline through pure creative practice in their free time. Taken together, these transformations signal a fundamental shift from creators who generate original content to correctors who continuously refine and optimize algorithmic outputs.
Finally we wish to acknowledge that this study also has some limitations. The fieldwork primarily focuses on creative teams associated with the Youku platform. While this allows for a detailed analysis of interactions between platforms and small teams, it pays little attention to independent creators operating outside platform systems. Additionally, AI technology is developing and the mechanisms of “technological monopoly” and “implicit exploitation” seen in this study may change as technologies advance and policies are updated. Future research could expand the sample to compare AI practices across different types of platforms, teams of various sizes, and policy contexts that may differ across regions or evolve over time. It could also monitor the evolution of creators’ bargaining strategies to see if creative workers can find effective ways to break free from current constraints over time. As AI continues to reshape the creative industries, understanding how power shifts through technology and impacts upon the art of animation is not only a key question for research but also a vital concern for the everyday lives of countless creative workers.
Footnotes
Acknowledgments
The author gratefully acknowledges the participating small- and medium-sized animation studios and the AI storyboarding development team at Youku, an Alibaba company. Sincere thanks are also extended to the anonymous reviewers and the editorial team of this special issue for their constructive and insightful comments, as well as to the interviewees who generously provided invaluable information for this study.
Ethical Considerations
This study was conducted following established ethical guidelines for research involving human participants. The research design received review and approval from the academic ethics committee at the institution affiliated with the first author. The second author contacted the Youku team through previous professional connections and conducted the research with organizational approval and awareness of relevant management. Before beginning the study, the second author informed the appropriate department and legal representatives about the research objectives, the data use scope, and publication plans. The organization was briefed on the nature of the research and consented to the use of materials under anonymity and without including sensitive information. Building on participant observation, the research team conducted semi-structured interviews. All participants received an information sheet beforehand that explained the study’s purpose, data use, anonymization procedures, and their right to withdraw at any time without consequences. Informed consent was obtained from all participants for the use of their responses in academic research and publications. To protect participants' privacy, all personally identifiable information and organizational identifiers have been anonymized or altered. Specifically, pseudonyms were assigned to all interview respondents and to the studied studios. Any additional details that could potentially identify individuals or organizations have been removed or modified. Interview recordings and transcripts are accessible only to the research team and are stored securely on encrypted devices. The second author’s background in the animation industry facilitated access and helped build rapport with participants. Throughout the research, reflexivity was maintained, and data were carefully reviewed and cross-checked to minimize potential bias. The research team did not receive any financial support or other forms of benefit from the research participants or their affiliated organizations that could constitute a conflict of interest. The funding supporting this research was obtained from independent academic sources and had no involvement in the research design, data collection, analysis, or interpretation.
Author Contributions
The first author was responsible for developing the theoretical framework and finalizing the manuscript, the second author conducted the fieldwork, collected data, and drafted the initial version, and the third and fourth authors contributed to the literature review and data analysis.
Funding
The authors received no financial support for the research, authorship, and/or publication of this article.
Declaration of Conflicting Interests
The authors declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.
Data Availability Statement
The datasets generated and analyzed during the current study are not publicly available due the fact that they constitute an excerpt of research in progress but are available from the corresponding author on reasonable request.
