Copy

AI generated videos are experiencing explosive growth and rapidly flooding major short video platforms, reshaping the pattern of information production and dissemination, leading to risks such as the proliferation of fake AIGC and information ecological pollution. Understanding and revealing how the signals provided by identification affect the interaction behavior between users and AI generated videos is the foundation for developing AI synthesized content identification methods and addressing deepfake content. This paper collects the data of AI generated videos in short video platforms such as WeChat video, Tiktok, Kwai, and Bilibili. Through regression analysis, it explores the impact of identification sources, AI generated short video content types and their interaction effects on user engagement behavior. The research conclusion of this article provides empirical evidence for the design of AI generated content identification for short video platforms, and provides reference for managers and service providers to implement and optimize AI generated composite content identification methods.