What kinds of political content thrive on TikTok during an election year? Our analysis of 51,680 political videos from the 2024 U.S. presidential cycle reveals that toxic and partisan content consistently attracts more user engagement—despite ongoing moderation efforts. Posts about immigration and election fraud, in particular, draw high levels of toxicity and attention. While Republican-leaning videos tend to reach more viewers, Democratic-leaning ones generate more active interactions like comments and shares. As TikTok becomes an important news source for many young voters, these patterns raise questions about how algorithmic curation might amplify divisive narratives and reshape political discourse.

• This study offers one of the first empirical examinations of how partisanship, political toxicity, and topical focus—such as immigration, racism, and election fraud—shaped user engagement with TikTok videos during the 2024 U.S. presidential election.
• The analysis was drawn from 51,680 political videos on TikTok, using models adjusted for feed ranking, user behavior (author, music, posting time), and platform engagement metrics.
• The majority of videos analyzed (77%) were explicitly partisan and were associated with approximately twice the engagement of nonpartisan content. Republican-leaning videos received more views, while Democratic-leaning ones showed more interactions—measured by total likes, comments, and shares.
• Toxic videos were associated with 2.3% more interactions. Partisan content also tended to show higher engagement, with Democratic-leaning toxic videos linked to even higher interactions.
• Racism, antisemitism, and election fraud were among the most toxic topics, with toxicity defined as rude or disrespectful language. Toxic videos on elections (+1.3%) and immigration (+3.5%) received higher engagement.
• Toxicity and engagement levels changed after major political events. Following Trump’s conviction, videos with severe toxicity and sexual attacks saw an approximate 2% surge in interactions.
• Captions alone were weak predictors of partisanship and toxicity, but transcripts of the audio from the videos improved alignment with manual labels (68.2%) and had significantly more (56.2%) toxic content. These results highlight the limitations of surface-level text features and the need for multimodal analysis in political content detection.