Silenced on social media: the gatekeeping functions of shadowbans in the American Twitterverse.

This article presents a large-scale audit study of shadowbanning on Twitter, a form of temporary, unannounced content moderation where user accounts or their posts are hidden or deprioritized. Analyzing a stratified random sample of approximately 25,000 U.S.-based Twitter accounts over multiple time points in 2020 and 2021, the study finds that shadowbans are relatively rare but more likely to affect accounts exhibiting bot-like behavior and less likely to impact verified or older accounts. Content featuring offensive language and political topics from both left- and right-leaning perspectives were more prone to downtiered replies, though these content-based effects showed temporal instability. The findings highlight Twitter’s tiered governance favoring socially influential users and underscore the fluid, reactive nature of algorithmic moderation, raising questions about transparency, fairness, and the ethical implications of platform governance in shaping online discourse.