Deepfakes have intensified fears about the power of synthetic media to reshape historical memory and blur the boundaries between truth and fabrication. This study compares how audiences process historical narratives presented by deepfaked figures and in the form of written text, focusing on differences in recall and perceived truth over time. In an online experiment, 332 U.S. participants read or watched first-person accounts by Albert Einstein or Marie Curie across two sessions one week apart. Written narratives produced higher recall and stronger perceived truth than deepfake videos, both immediately and one week later. Also, Need for Cognition predicted greater perceived truth across conditions, suggesting that analytic engagement rather than technological realism drives belief in mediated representations. Although perceived truth declined over time, the drop was steeper for deepfake messages. Age also emerged as a significant predictor of some measures of perceived truth and recall. Pairwise comparisons between age groups showed progressive increases with participants aged 18–29 displaying the weakest recall and truth ratings, whereas those aged 30 and above maintained stronger and more stable judgments over time, independent of analytic motivation. These findings indicate that analytically engaged audiences are resilient to synthetic realism, limiting the potential for deepfakes to distort historical understanding. • Deepfake videos produced lower recall accuracy than equivalent text narratives. • Written accounts were judged as more truthful than their deepfake counterparts. • Deepfakes prompted greater skepticism despite their audiovisual realism. • Memory and perceived truth decayed faster for deepfakes than for text messages. • Text modality fostered deeper processing and more stable truth judgments than deepfakes. • These findings challenge assumptions that audiovisual realism in deepfakes enhances credibility. • Deepfaking the Past: Memory and Perceived Truth of Resurrected Historical Figures.
Building similarity graph...
Analyzing shared references across papers
Loading...
María T. Soto-Sanfiel
Gina Junhan Fu
Computers in Human Behavior
National University of Singapore
Building similarity graph...
Analyzing shared references across papers
Loading...
Soto-Sanfiel et al. (Wed,) studied this question.
www.synapsesocial.com/papers/69cf5eee5a333a821460dafc — DOI: https://doi.org/10.1016/j.chb.2026.109008
Synapse has enriched 5 closely related papers on similar clinical questions. Consider them for comparative context: