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

Abstract

We present the first quantitative study of deepfake activity during the 2024 U.S. presidential election by analyzing a novel dataset of 231 deepfakes (images, videos, and audio clips) across social media during the May-December 2024 election period. Our comprehensive statistical analysis examines five research questions: (1) whether deepfake publication spikes occur around key election events (KEEs), (2) whether there is a temporal relationship between deepfakes and KEEs (before, during, or after), (3) which specific types of KEEs trigger deepfake spikes, (4) whether KEEs boost engagement with deepfakes, and (5) the differential impact of various KEEs on deepfake engagement. Our findings reveal that spikes in deepfake activity preceded KEEs, and engagement with deepfakes (e.g., likes, comments) surged during pre-KEE time windows. Our curated dataset offers researchers a valuable resource to study the impact of synthetic media in political contexts, while our findings provide valuable advice for policymakers and social platforms to develop appropriate measures to counter potential malign deepfakes before future elections.