Abstract: [Purpose/Significance] The use of anthropomorphic digital human images in social media has raised concerns about the authenticity and trustworthiness of the content. Exploring users' ability to identify digital human images is crucial for the management of AI-generated content in social media. [Design/Methodology] 60 participants were recruited for an eye-tracking experiment to validate the influencing factors of digital human image identification and the role of perceived authenticity in content trustworthiness. [Findings/Conclusion] Users exhibit limited ability to identify digital human images on social media. Visual characteristics, such as background, attractiveness and gender of characters in images, significantly influence users' perception of the authenticity of digital human images. Meanwhile, perceived authenticity significantly improves the trustworthiness of graphic posts. Users' interest plays a significant moderating role in the relationship between authenticity perception and content trustworthiness. [Originality/Value] This study constructs a model of factors influencing digital human image identification and improves the identification mechanism for AI-generated content and the explanatory framework for human-machine trust, which provide a reference for the use and management of digital human images.

Keywords: Authenticity perception, Digital human, Social media, Physical attractiveness, Human-machine trust