<p>We advance and test a theory-driven, multi-mechanism model in which platform-related attentional capture, dependence on social validation, and anticipatory worry jointly <i>account for</i> social media’s association with emotional distress. In a community sample of adults (<i>n</i> = 444), ages 18–56&#xa0;years, 56.8% female, bootstrapped mediation and out-of-sample interpretability analyses converged on attention and worry as the dominant correlates, with age-differentiated weights. Multivariable models indicated that attention difficulties, validation seeking, and worry were each positively associated with distress, whereas age was negatively associated. In the full model a small residual direct association between social media and distress remained. Mediation analyses using nonparametric bootstrapping supported significant indirect mediated associations linking social media and distress via attention, worry, and validation seeking. Moderation tests were not supported. Out-of-sample prediction with five-fold cross-validated models favored regularized linear models (Ridge <i>R</i>² ≈ <i>.</i>488), and model-interpretability methods converged on attention and worry as the most influential correlates of distress. Overall, findings suggest that social media’s link to emotional distress is primarily indirect, with indirect associations via attention dysregulation, validation dependence, and worry, with a small residual direct association. Interventions targeting these potential mechanisms may help reduce the psychological burden of heavy social media use and promote healthier digital habits.</p> Graphical Abstract <p></p>