Nineteen Years of ASMR on YouTube: A Multilingual, Theme-Level Analysis of 42,268 Videos

Alam, M. S., Bazilinskyy, P.

Submitted. (2026)

ABSTRACT ASMR videos have become a major genre on online platforms, yet their large scale characteristics remain underexplored. Using the YouTube Data API and a pytubefix workflow, we assemble a dataset of 42,268 ASMR videos from 8,587 channels (2008 to 2026, 34 languages) enriched with duration, views, likes, inferred language, theme flags, and lemmatised title and description text. English dominates (76.94% of videos), followed by Korean, Japanese, Spanish, Dutch, and Portuguese. Across the corpus, the mean growth is 1,786.69 views per day and the duration analysis shows that short videos (under 10 minutes) average 3,435.52 views per day versus 1,005.91 for 10 to 30 minute content, while very long (over 180 minutes) videos reach 2,481.93 views per day. Theme detection indicates that drive themed content (17.14%), sleep related content (17.13%), and visual trigger content (15.54%) are particularly prevalent, with whisper (10.83%) and binaural videos (8.55%) also common. K means clustering on multimodal text, language, and engagement features, visualised with tSNE, yields 11 content clusters (1 to 29,321 videos) and a small set of extremely high growth outliers.