When Seeing Is Not Believing: Reassessing the Persuasive Power of Deepfakes as Opposed to Text-Based First-Person Narratives of Historical Figures

Amid growing ethical concerns over the viral spread of deepfakes, existing literature suggests that their sensory richness and enhanced realism evoke complex, sometimes contradictory, audience reactions. This study examines the persuasive potential of deepfakes by comparing audience responses to identical resurrection narratives presented in deepfake video format versus text-based storytelling. Participants (N = 813, Mage = 37.28, SDage = 12.46) were exposed to either a deepfake video or a textual narrative featuring a historical scientist (Albert Einstein or Marie Curie) recounting their life story and then completed scales measuring variables related to narrative persuasion. Results revealed a “reverse modality effect,” with text narratives generating significantly higher transportation than deepfakes. Structural equation modeling showed that this effect extended to downstream outcomes, including perceived realism, message credibility and evaluation, character identification, enjoyment, and sharing intention. Significant indirect effects of modality emerged across all persuasion variables. Multigroup analysis revealed partial scalar invariance, with Einstein eliciting stronger identification and enjoyment than Curie. These findings suggest that deepfake resurrection narratives engage distinct psychological mechanisms that may disrupt the persuasive advantages typically attributed to the richness of audiovisual media and to deepfakes in particular. Moreover, the impact of resurrection deepfakes appears to be conditioned not only by their technical sophistication but also by the cultural resonance of the figures they revive, raising implications for ethical and effective use in educational, cultural, and prosocial domains.