Online learning engagement plays a critical role in shaping learners' academic outcomes. Few studies have systematically explored the roles and mechanisms of learners' beliefs in online learning. Based on the integrated self-theories and the control-value theory of achievement emotions, this empirical research examines implicit theories of intelligence, learners' beliefs about the malleability of their intelligence, and explores whether these beliefs impact online learning engagement through the mediating role of achievement emotions during online learning. In order to preliminarily explore the potential relationship model among the factors, Study 1 adopts a cross-sectional design to explore the relationships among implicit theories of intelligence, achievement emotions, and online learning engagement among Chinese university students. 405 Chinese undergraduate participants (female = 323) were recruited from 113 universities across four economic regions of mainland China. Study 2 further utilizes an experimental approach, manipulating implicit theories of intelligence to replicate the findings from Study 1 in an immediate online learning context. In Study 2, a single-factor between-subjects experimental design was adopted to recruit 131 Chinese university students as participants (102 females), who were then randomly assigned to either the incremental theory group (n = 65) or the entity mindset group (n = 65). The collective results from both studies underscore a consistent pattern: boredom emerges as a significant mediator, significantly fully mediating the negative impact of entity theory on three online learning engagement. The direct effects of implicit theory of intelligence on online learning engagement are not significant. The mediating roles of enjoyment, anxiety, and hopelessness were found to be non-significant. This study illuminates complex relationships among belief factors, academic emotions, and online learning engagement in college students. It integrates theories and adopts a systematic approach combining cross-sectional design with experimentation, laying a foundation for future research, also provides scientific evidence for online educational practices.