A Nonpartisan Study of Visual Deepfake Activity and Engagement Around the 2024 US Presidential Election

The 2024 U.S. Presidential Election Deepfakes Dataset (USPED) was created to address a research gap in the study of deepfakes in electoral contexts. While previous studies have documented instances of synthetic media in various elections worldwide, none offered both a comprehensive, publicly available dataset and rigorous quantitative analysis of deepfake dynamics. This dataset was specifically designed to investigate the relationship between Key Election Events (KEEs) and deepfake activity through statistical analysis of publication patterns and engagement metrics. By documenting 231 carefully curated deepfakes (169 images, 38 videos, and 24 audios) from the 2024 U.S. Presidential Election cycle, this resource enables researchers to empirically examine how synthetic media interacts with significant political moments. The dataset will be made available to academics who agree to an ethical usage policy, supporting future research on the potential impact of deepfakes on democratic processes.

The dataset is associated with the work "A Nonpartisan Study of Visual Deepfake Activity and Engagement Around the 2024 US Presidential Election". Please refer to the datasheet for further information.