Algorithmic identity performance: self-censorship and the politics of visibility on Chinese short-video platforms

Short-video creators on China’s Douyin face a visibility dilemma: algorithms distribute attention and mediate income opportunities, while moderation and reporting make acceptable expression uncertain. Existing scholarship has extensively examined censorship in China, but less is known about how creators translate layered constraints into routine production decisions and identity work. Based on 27 semi-structured interviews with Douyin content creators, we conceptualize self-censorship as an iterative practice of managing visibility, legitimacy, and safety. Creators read audience and platform cues, anticipate sanctions, and adapt through pre-editing, topic substitution, linguistic recoding, and selective withdrawal, with consequences for discourse diversity and civic expression. We develop Algorithmic Identity Performance (AIP), a process framework showing how creators detect governance cues, interpret them through practical theories, and perform platform-compatible selves via feedback loops driven by visibility incentives. AIP demonstrates how self-censorship extends beyond speech or silence to sustained identity performance.