Synthetic Politics: Prevalence, Spreaders, and Emotional Reception of AI-Generated Political Images on X

TL;DR

This work addresses the scale, spread, and reception of AI-generated political imagery on X during the 2024 U.S. election. It introduces a GPT-4o-based detection framework with rigorous validation to quantify prevalence ($ ext{$12.33\%$}$) and diffusion concentration ($ ext{$80\%$}$ of AI shares from about $10\%$ of spreaders) and identifies 128 AIGC superspreaders using an $h$-index approach. It then profiles these superspreaders via AI and bot scores, partisan signals, and premium status, and analyzes audience reactions through emotion and toxicity analyses of 16,832 AI-image comments and 27,848 non-AI-image comments. The findings reveal an unequal yet widespread presence of AIGC, a notable association with bots and premium accounts, heterogeneous sharing with extreme cases, and a more positive, less toxic response to AI-generated images, offering implications for platform governance and public discourse research. These insights inform policy design, cross-platform comparisons, and ongoing efforts to understand generative AI’s impact on political communication and online sociopolitical environments.

Abstract

Despite widespread concerns about the risks of AI-generated content (AIGC) to the integrity of social media discourse, little is known about its scale and scope, the actors responsible for its dissemination online, and the user responses it elicits. In this work, we measure and characterize the prevalence, spreaders, and emotional reception of AI-generated political images. Analyzing a large-scale dataset from Twitter/X related to the 2024 U.S. Presidential Election, we find that approximately 12% of shared images are detected as AI-generated, and around 10% of users are responsible for sharing 80% of AI-generated images. AIGC superspreaders--defined as the users who not only share a high volume of AI-generated images but also receive substantial engagement through retweets--are more likely to be X Premium subscribers, have a right-leaning orientation, and exhibit automated behavior. Their profiles contain a higher proportion of AI-generated images than non-superspreaders, and some engage in extreme levels of AIGC sharing. Moreover, superspreaders' AI image tweets elicit more positive and less toxic responses than their non-AI image tweets. This study serves as one of the first steps toward understanding the role generative AI plays in shaping online socio-political environments and offers implications for platform governance.

Figures (6)

• Figure 1: Pipeline of the detection and validation of AI-generated images.
• Figure 2: Cumulative distribution function (CDF) of the total number of AI-generated images shared by image spreaders.
• Figure 3: AI scores (left) and bot scores (right) across user groups. According to the Mann-Whitney U test, Superspreaders exhibit significantly higher AI scores than Non-Superspreaders ($p$<.001), and significantly higher bot scores ($p$<.01).
• Figure 4: Comparison of emotion scores across AI image tweets vs. non-AI image tweets, both shared by AIGC superspreaders. We apply the Mann-Whitney U test to compare the two tweet groups across emotion categories. All categories shown are statistically significant (* $p$$<$ .05, ** $p$$<$ .01, *** $p$$<$ .001); The emotion surprise is not shown as it is not significant.
• Figure 5: Comparison of toxicity scores between AI and non-AI image tweets shared by AIGC superspreaders. The Mann-Whitney U test indicates a significant difference ($p$<.001)