Purpose <p>This randomized controlled trial (RCT) evaluated the efficacy of a novel stepped-care digital game intervention for improving emotion recognition in autistic children.</p> Methods <p>Thirty-six autistic children (aged 3–10 years) were randomly assigned to an experimental group (<i>N</i> = 18) or a control group (<i>N</i> = 18). The intervention, grounded in the Theory of Mind framework, comprised three hierarchically structured games designed to progressively train emotion recognition. Pretest and posttest including questionnaires and emotion recognition tests were administered before and after the intervention.</p> Results <p>Results indicated that the experimental group showed significant improvements in accuracy on the emotion recognition tests, whereas the control group did not. A significant interaction was found for scores on the autism treatment evaluation checklist (ATEC), suggesting differential change patterns across groups. Parenting stress increased significantly in the control group but remained stable in the experimental group.</p> Conclusion <p>These findings demonstrated that this stepped-care digital game intervention can effectively enhance emotion recognition in autistic children and may bring benefits for their core traits and parenting stress.</p>