Perceived AIGC saturation and AI content fatigue on short-video platforms: An appraisal-based structural equation modeling study

Abstract

Artificial intelligence-generated content (AIGC) is increasingly embedded in short-video platforms, where algorithmic recommendation systems expose users to high volumes of rapidly produced synthetic media. Although prior research has examined generative AI from the perspectives of creativity, productivity, and technology adoption, less is known about how ordinary users evaluate repeated exposure to AI-generated short videos. This study develops and tests an appraisal-based model of AI content fatigue, distinguishing perceived AIGC saturation from users’ appraisals of low-quality prevalence, narrative homogenization, and content overload. A cross-sectional online survey was conducted among Chinese short-video users in May 2026. After excluding invalid responses, 513 valid questionnaires were analyzed using confirmatory factor analysis and structural equation modeling. The adjusted structural model showed acceptable fit to the data, χ²(1321) = 1991.541, χ²/df = 1.508, CFI = .946, TLI = .929, IFI = .947, and RMSEA = .031. Perceived AIGC saturation was positively associated with perceived low-quality prevalence, perceived narrative homogenization, and perceived content overload. Perceived low-quality prevalence had the largest estimated coefficient, followed by perceived content overload, whereas perceived narrative homogenization was not supported under bootstrap inference. AI content fatigue was negatively associated with satisfaction and continued viewing intention. Bootstrap analyses indicated 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, whereas those through perceived narrative homogenization were not. The findings indicate that fatigue in AIGC-rich short-video environments is not simply a response to AI-generated content itself. It is more closely related to users’ appraisals of content quality, repetitiveness, and cognitive burden. The study contributes to social media fatigue research by conceptualizing AI content fatigue as a content-specific evaluative state and offers implications for platform governance, AIGC labeling, recommendation diversity, and quality-sensitive moderation.