Assessing the perceived credibility of deepfakes: The impact of system-generated cues and video characteristics

The adverse effects of deepfakes are becoming apparent; however, less is known about what affects individuals’ credibility judgments concerning deepfakes. This article conducted two randomized controlled 2 × 2 × 2 experiments ( N = 518) to explore the influence of heuristic cues (including the number of followers, popularity, and description as variables) generated by the system and video characteristics (including definition, duration, and editing as variables). The results suggest that the number of the source’s followers and the video’s popularity are positively associated with perceived credibility. Moreover, high-definition deepfakes are more efficient at deceiving viewers. Interestingly, an interaction effect indicates that it is not easy for users to detect editing traces in relatively long videos (greater than 30 s). Our study is an initial step to better understand how individuals assess video credibility and lay the grounds for approaches to combat the negative effects of deepfakes.