Perceived AIGC saturation is eroding short-video engagement
Artificial intelligence-generated content is no longer a novelty inside short-video feeds. On crowded platforms built around rapid scrolling and algorithmic recommendation, repeated exposure to synthetic clips is beginning to feel less like innovation and more like churn.
A cross-sectional survey of Chinese short-video users found that perceived AIGC saturation was associated with a distinct form of AI content fatigue. The study moved beyond the familiar question of whether people like generative tools and instead asked how ordinary viewers judge repeated AI-made videos when they encounter them again and again in the feed.
The signal was strongest where users perceived low quality. Content that looked mass-produced, thin, or disposable appeared to drive fatigue more powerfully than any abstract discomfort with AI itself.
Appraisal, not technology alone, shaped user fatigue
The research used an appraisal-based structural equation model to separate perceived AIGC saturation from three user judgments: perceived low-quality prevalence, perceived narrative homogenization, and perceived content overload. Each of those appraisals captures a different way audiences can experience AI-heavy feeds. One is about craft. One is about sameness. One is about mental strain.
That separation matters because AI content fatigue did not emerge as a single blunt reaction to machine-made media. It emerged through the viewer’s own evaluation of what the platform is serving, how repetitive it feels, and how much effort it takes to keep sorting worthwhile clips from clutter.
The sample came from a cross-sectional online survey conducted in May 2026 among Chinese short-video users. After invalid responses were removed, 513 questionnaires remained for analysis. The study then applied confirmatory factor analysis and structural equation modeling to test the proposed pathways.
Model fit was acceptable across standard indices: χ²(1321) = 1991.541, χ²/df = 1.508, CFI = .946, TLI = .929, IFI = .947, and RMSEA = .031. Those values support the internal coherence of the appraisal model rather than a loose descriptive association.
Low-quality prevalence carried the heaviest statistical weight
Perceived AIGC saturation was positively associated with perceived low-quality prevalence, perceived narrative homogenization, and perceived content overload. Among those paths, low-quality prevalence carried the largest estimated coefficient, followed by content overload.
Perceived narrative homogenization, by contrast, did not hold under bootstrap inference. In practical terms, users were not merely reacting to sameness in story structure. They were more sensitive to a broader sense that too many AI-generated videos looked weak, lacked substance, or demanded extra effort to filter out.
The hierarchy is clinically simple in its own way. People fatigue first when content feels cheap, then when it becomes cognitively tiring.
This pattern suggests a platform environment in which quality signals may be doing more damage than repetition alone. A user can tolerate a familiar format if the clip still feels fresh, useful, or entertaining. When a feed becomes saturated with low-value synthetic material, the sense of overload may sharpen quickly and suppress engagement.
Fatigue translated into lower satisfaction and weaker return interest
AI content fatigue was negatively associated with satisfaction and continued viewing intention. The study also found selected path-specific indirect associations from perceived AIGC saturation to AI content fatigue, satisfaction, and continued viewing intention.
Indirect pathways through perceived low-quality prevalence and perceived content overload were supported. The route through perceived narrative homogenization was not.
This is the core mechanism the analysis brings into focus. Saturation alone was not enough to explain the outcome. Users’ appraisals turned saturation into fatigue, and fatigue then fed into lower satisfaction and a weaker desire to keep watching. In algorithmic environments, where the next clip is always one swipe away, even small degradations in satisfaction can change how long a viewer stays inside the platform.
The result also fits the broader logic of social media fatigue research. Overload, exhaustion, and disengagement tend to follow when users face too much of a stimulus that requires repeated assessment. Here, the stimulus is not general social content but AI-generated short video content with uncertain quality and unclear originality.
Platform governance now has a quality problem, not just a volume problem
The findings carry direct implications for short-video governance. Labeling AIGC may help users recognize synthetic media, but disclosure alone will not solve fatigue if the feed remains saturated with low-quality output. Diversity in recommendation systems may matter because repetitive exposure can intensify the burden of sorting, judging, and dismissing weak content. Quality-sensitive moderation may also be needed when platforms optimize for volume and velocity rather than editorial value.
The study’s conceptual contribution is equally clear. It frames AI content fatigue as a content-specific evaluative state rather than a generic reaction to emerging technology. That framing is useful because it shifts attention away from the fact that content is AI-generated and toward how that content is experienced in daily use.
In a short-video environment, users are not watching in a vacuum. They are navigating an endless stream shaped by recommender systems, template production, and rapid turnover. When those systems amplify low-quality synthetic clips, fatigue can become part of the user experience itself.
What the analysis can and cannot settle
The evidence comes from a cross-sectional survey, so it cannot establish temporal causation. The data still show a coherent pathway from perceived AIGC saturation to fatigue-related outcomes, but they do not track how those perceptions evolve over time or whether users eventually adapt to heavier AI exposure.
The study also relies on self-reported appraisals rather than objective measures of platform content quality or algorithmic ranking. Even so, the distinction is meaningful because perceived quality, overload, and repetition are the psychological variables most likely to determine whether a user keeps scrolling or leaves the feed.
The article leaves one practical question open: if platforms increase AI output without improving quality control, will users continue to disengage in larger numbers, or will they simply recalibrate what they are willing to tolerate? The survey does not answer that on its own, but it makes the direction of risk difficult to ignore.