From AI labels to trust: cognitive pathways linking authenticity judgments, risk perception, and trust in AI-generated videos
Abstract
Background The rapid advancement of artificial intelligence-generated content (AIGC) has raised concerns regarding the authenticity and trustworthiness of synthetic media. Although AI disclosure labels are increasingly used as transparency mechanisms, the cognitive processes through which these cues influence trust formation remain unclear. Objective This study aimed to develop and test a cognitive pathway model explaining how AI disclosure labels, authenticity judgments, perceived risk, cognitive processing styles, and literacy factors shape trust toward AI-generated videos. Methods A 2 × 2 between-subjects experiment was conducted with 617 participants randomly assigned to four conditions involving AI-generated instructional videos with different AI label and video quality combinations. Participants evaluated authenticity, systematic and heuristic processing, trust, perceived risk, adoption intention, sharing intention, AI literacy, and media literacy. Regression, bootstrapped mediation, and moderation analyses were performed. Results AI disclosure labels significantly reduced perceived authenticity but did not directly affect trust, adoption, sharing intention, or perceived risk. Trust formation occurred through cognitive pathways: systematic processing increased trust, whereas heuristic processing reduced trust. Authenticity enhanced trust and lowered perceived risk, with perceived risk mediating the authenticity–trust relationship. Trust positively predicted adoption and sharing intentions, while AI and media literacy moderated selected relationships. Conclusion Trust in AI-generated videos was not directly determined by transparency labels alone. Instead, the observed associations among authenticity judgments, perceived risk, information processing, and trust were consistent with a multi-stage cognitive interpretation, although the cross-sectional post-exposure measurement of these constructs does not establish their temporal or causal sequence.