PurposeThe purpose of this study is to unpack the antecedents and consequences of clickbait prevalence in online media at two different levels, namely, (1) Headline-level: what characteristics of clickbait headlines attract user clicks and (2) Publisher-level: what happens to publishers who create clickbait on a prolonged basis.Design/methodology/approachTo test the proposed conjectures, the authors collected longitudinal data in collaboration with a leading company that operates more than 500 WeChat official accounts in China. This study proposed a text mining framework to extract and quantify clickbait rhetorical features (i.e. hyperbole, insinuation, puzzle, and visual rhetoric). Econometric analysis was employed for empirical validation.FindingsThe findings revealed that (1) hyperbole, insinuation, and visual rhetoric entice users to click the baited headlines, (2) there is an inverted U-shaped relationship between the number of clickbait headlines posted by a publisher and its visit traffic, and (3) this non-linear relationship is moderated by the publisher's age.Research limitations/implicationsThis research contributes to current literature on clickbait detection and clickbait consequences. Future studies can design more sophisticated methods for extracting rhetorical characteristics and implement in different languages.Practical implicationsThe findings could aid online media publishers to design attractive headlines and develop clickbait strategies to avoid user churn, and help managers enact appropriate regulations and policies to control clickbait prevalence.Originality/valueThe authors propose a novel text mining framework to quantify rhetoric embedded in clickbait. This study empirically investigates antecedents and consequences of clickbait prevalence through an exploratory study of WeChat in China.