More Clickbait, But Fewer Shares? Unpacking the Relationship Between Multimodal Sensationalism and Online User Engagement

This study examined multimodal sensationalism by analyzing how textual and visual sensationalist features embedded in video headlines and thumbnails relate to user engagement. Using a hybrid human-large language model content analysis, we identified 15 sensationalist features across 1,100 trending videos from Douyin. The results show that legacy news outlets used fewer textual sensationalist elements, particularly Internet slang, hashtags, and positive emotions, than non-journalistic producers. Additionally, visual sensationalist features exhibited an inverted U-shaped relationship with user engagement. These findings advance understanding of the nonlinear and modality-specific effects of sensationalism and offer practical insights into the optimal use of multimodal sensationalism on social media.