Abstract: Consumer online engagement behavior (COEB) is critical to effective social marketing. To reevaluate a number of factors that influence COEB, such as informativeness and positivity, based on the Elaboration Likelihood Model and Impression Management, this study adopts a newly designed natural language processing algorithm-an unsupervised domain adaptation system for content informativeness prediction-and builds a unified theoretical framework using data from Sina Weibo's beauty microblogging community. This differs from the existing literature, which focuses on minor features of online content. The results show that:(1) content informativeness has a non-linear positive impact on COEB; (2) content positivity has a positive impact on COEB and positively moderates the relationship between informativeness and COEB; and (3) informativeness and positivity have differentiated impacts on COEB. This study not only improves the theoretical understanding of COEB but also offers practical guidance for firms on how to effectively communicate with consumers on social platforms.

Key words: social media, social media marketing, content informativeness, online engagement behavior