Dynamic News Visibility Across Media Sources: Evidence from a Chinese Algorithmic Platform

This study examines how recommendation systems organize differential news visibility on Douyin, one of China’s largest short-video platforms. Focusing on the visibility dynamics between government-affiliated media and non-government-affiliated media, the study adopts an agent-based testing (ABT) approach to audit how Douyin’s recommendation system allocates exposure under different stages of user interaction and preference formation. The findings suggest that government-affiliated media receive greater visibility during the early stages of recommendation, when user behavioral data remain limited. However, as recommendation systems accumulate interaction signals over time, non-government-affiliated media gradually gain visibility, particularly under conditions of sustained user engagement. At the same time, the recommendation system does not respond symmetrically to different forms of user preference, revealing varying degrees of algorithmic responsiveness in the visibility allocation process. By examining how recommendation systems negotiate commercial, institutional, and behavioral priorities in news distribution, this study contributes to research on algorithmic visibility, platform governance, and algorithm audit.