<p>Engagement is a well-known challenge in digital health interventions (DHIs); therefore, being able to accurately distinguish those who do versus do not engage with these interventions is of critical importance. However, efforts to predict DHI engagement have been limited for two main reasons: (1) the narrow range of generic predictor variables explored and (2) failure to consider how putative predictors may interact to influence engagement. Self-regulated learning (SRL) is the process of actively controlling one’s cognitions, motivations, and behaviours to improve learning outcomes, and it provides a novel, albeit highly relevant, framework for exploring DHI engagement, particularly for programmes that deliver content as self-guided psychoeducation. Nevertheless, its predictive value has yet to be empirically tested. In this study, we apply decision tree analysis (a machine-learning-based approach) to explore whether key SRL variables interact to predict higher or lower engagement in a randomised controlled trial of a smartphone app intervention for individuals with recurrent binge eating (<i>n</i> = 474). Decision trees revealed complex interactions between several SRL variables, including social support, self-efficacy, self-control, task interest, and motivation. Importantly, the analysis identified potential trade-off and amplification effects among these variables that explained individual differences across two key engagement outcomes (i.e. number intervention modules completed and skill activities unlocked). These findings highlight the complexity of DHI engagement, suggesting that it is not shaped by any one predictor in isolation but by the complex interplay of multiple underlying psychological processes. Nevertheless, given the exploratory nature of this analytical approach, further testing is still needed to validate this pattern of findings. </p><p>Clinical Trial Number. This clinical trial is registered with the Australian New Zealand Clinical Trials Registry (ACTRN12624000158561).</p>