Generative artificial intelligence, misinformation and political polarization: a content analysis of deepfake propaganda during the 2024 US presidential election

Transformative digital platforms, including generative artificial intelligence (GenAI), have accelerated the production and circulation of misinformation in contemporary political communication. While the creation of false or misleading content is a significant concern, the perceived credibility of such material and the ways it is taken up within online discourse generate negative externalities that extend beyond individual users. This study examines how AI-driven deepfake artifacts related to the 2024 U.S. presidential election are framed and discussed within polarized political discourse on the social media platform X (formerly Twitter). Using qualitative content analysis, the study analyzes AI-generated deepfake images and videos alongside user comments responding to these artifacts. The findings show that deepfake content is embedded within highly polarized discourse characterized by emotional intensity, partisan alignment, and antagonistic representations of political figures. Recurring themes include emotional polarization, political satire and humor, and manipulative framing strategies that structure how candidates and political issues are portrayed in deepfake artifacts and subsequent user responses. Rather than assessing behavioral effects or audience influence, the analysis highlights how deepfakes circulate within online environments that reflect and reproduce polarized narratives. Based on these patterns, the study proposes the Deepfake-Driven Framing and Polarization Framework (DDFPF) to conceptualize how different types of deepfake images and videos correspond to distinct patterns of framing and commenter engagement in polarized political contexts.