The growing influence of digital technologies is fundamentally changing communication patterns on short video platforms. Stable social communities defined by fixed affiliations and institutionalised structures are being replaced by ‘light communities’. These are connected by loose forms of participation. This paper analyses User-Generated Content (UGC) comments on the Douyin and TikTok platforms. To this end, an innovative, corpus-based AI analysis framework (UTA-AI model) was developed that systematically examines platform-specific characteristics of user-generated comments along three analytical coupling levels: the user-topic, user-affect, and topic-affect relationships. The framework methodically integrates quantitative procedures (including automatic sentiment analysis using the Baidu AI API) with qualitative interpretation. The aim of this three-dimensional approach is to understand the emergence of and interaction within light communities. In addition, differences between the two platforms are identified. The results gained provide an insight into the paradigmatic change in media texts in the AI age and its influence on user behaviour. In the future, the results of this work can provide a data-driven basis for adaptive influencer marketing strategies and better targeted advertising design.