What’s on Their For You Page? A Large-Scale Computational Approach to Analyzing Adolescents’ TikTok Archives Through Hashtag Topic Modeling

Abstract

Short-form video platforms like TikTok play a central role in adolescents’ lives, yet little is known about the actual content they engage with. This study presents a large-scale, computational approach to analyzing adolescents’ TikTok feeds using hashtag-based topic modeling applied to user-donated archives from 102 Dutch adolescents (ages 15–19). Using BERTopic, we identify 304 coherent topics, organize them into 32 higher-order content categories, and validate both through a structured human validation study. Analyses reveal that adolescents’ feeds are dominated by entertainment and leisure content, with ‘Media productions,’ ‘Sports,’ ‘Style and appearance,’ ‘Hobbies,’ and ‘Gaming’ as the most prevalent categories. At the same time, there is marked individual variation: niche areas such as K-pop, anime, and specific fandoms are central to some users’ feeds. Latent profile analysis identifies six distinct, gendered content-consumption profiles. Together, the findings shed light on adolescents’ TikTok exposure, revealing predominantly benign but highly personalized media diets in which appearance-focused content stands out as a potential risk domain. In all, our approach offers a scalable, efficient way to map short-form video content. Linking these maps to individual users’ media diets opens new possibilities for examining how different content patterns relate to user experiences and outcomes.