Visual richness and consumer engagement in short-form tourism videos: a computational analysis of additive and compositional effects

Purpose Despite the widespread use of short-form videos in tourism marketing, it remains unclear why videos using the same medium vary substantially in their ability to stimulate travel intention and recommendation behavior. This study aims to address this gap by examining how content-level visual richness in short-form tourism videos shapes different forms of viewer engagement, thereby advancing understanding of visual persuasion in short-form tourism communication. Design/methodology/approach Drawing on media richness theory (MRT), this study conceptualizes visual richness as a content-level property of short-form tourism videos and examines the richness of scenery visuals and portrait visuals respectively, as well as their compositional ratios. Based on data of 1,941 Douyin travel videos and 676,500 user reviews, keyframe extraction and deep-learning classification are applied to quantify visual features, which are further linked to travel intention and recommendation behavior through mixed-effects negative binomial models. Findings Higher visual richness enhances both outcomes, while both scenery visual richness and portrait visual richness contribute positively to engagement, with scenery-based cues showing slightly larger coefficients. Recommendation behavior displays a U-shaped response to visual cue composition, such that videos with a clear visual emphasis – either scenery-dominant or portrait-dominant – are more likely to be recommended than mixed compositions. On the other hand, portrait-heavy videos weaken travel intention. Originality/value This study extends MRT by conceptualizing visual richness in short-form videos as both an additive and a compositional property. It shows that effects of visual richness are not uniformly increasing, but bounded and configuration-dependent. In particular, although scenery and portrait cues are each positively associated with engagement, moderately mixed visual compositions may be less effective in stimulating recommendation than videos with a clear visual emphasis.