When bots act like us: the role of social media accounts' human-likeness in opinion polarization

Purpose As social bots penetrate social media, social media accounts with different human-likeness interact with each other, with opinion polarization as one of the results. This study investigates the impact of human-likeness of social media accounts on opinion polarization on social media. Design/methodology/approach Drawing on the social validation principle and its intersections with the optimal distinctiveness theory as the theoretical foundation, this study applies empirical analysis with a dataset of 2,034,201 replies from Weibo to examine the proposed model. Findings An inverted U-shaped relationship between human-likeness of social media accounts and opinion polarization in replies mediated by their social validation and moderated by information valence. Originality/value Our findings advance understanding of human–bot interaction dynamics, highlighting how bot human-likeness relates to opinion polarization at the community level. This study provides actionable measures for stakeholders: bot developers should carefully calibrate human-likeness, platforms should implement graded indicators on human-likeness of social media accounts and users must critically evaluate information sources to mitigate polarization.