Recommending the state: How social media algorithms curate state-created content in China

Authoritarian governments have increasingly expanded their social media presence, producing massive content to shape public opinion and behavior. However, less is known about how social media algorithms curate such content and serve authoritarian political goals. This study theorizes algorithmic promotional curation, whereby recommendation algorithms systematically amplify state-created content, and empirically examines this curation through analyzing 119,064 trending and recommended videos from Bilibili, one of China’s largest video-sharing platforms. Using descriptive analyses, regression models, and Markov chain simulations, we find that state-created content is disproportionately amplified through video recommendations associated with state-created trending videos, and that state-affiliated accounts exhibit strong self-reinforcement. Yet this algorithmic promotional curation is not uniform across content categories, with stronger amplification among state-created news and politics than other content. These findings demonstrate how recommendation systems may subtly serve authoritarian goals, advance a multilateral understanding of algorithmic curation, and extend authoritarian information control frameworks beyond censorship and propaganda.