In contemporary China, amidst rapid socioeconomic transformation, the frequent emergence of collective action profoundly reflects and reshapes societal dynamics, underscoring the critical importance of in-depth research into this phenomenon. Previous studies have primarily leveraged media data or social media texts and behaviors to predict participation in collective action. This paper aims to empirically analyze collective action in China by examining textual data from the Weibo platform. Our findings indicate that goal-oriented “acquire” language positively predicts the occurrence of disruptive and violent collective actions. Conversely, analytical “insight” language demonstrates a significant negative predictive power for violent actions. Notably, the predictive efficacy of political and cultural terms diminishes once structural variables are controlled for. This study reveals the significant influence of actor motivations and cognitive depth within online discourse on participation in collective action, thereby offering new empirical evidence and avenues for research into collective action in the digital age.