Purpose: Previous cross-sectional research suggests self-compassion may mediate associations between autistic traits and mental health in autistic and non-autistic adults. However, no research to date has examined these relationships longitudinally. In this study, we used a cross-lagged panel analysis to examine correlations over time between autistic traits, self-compassion, and anxiety/depression. Methods: Participants were from the UK and included autistic (n = 228 at T1, n = 156 at T2, and n = 165 at T3) and non-autistic adults (n = 228 at T1, n = 122 at T2, and n = 124 at T3) without intellectual disability. Participants were recruited through an online participation platform and completed demographics, the Autism Spectrum Quotient (AQ), the Self-Compassion Scale (SCS), and the Hospital Anxiety and Depression Scale (HADS) at baseline (T1), 6 months (T2), and 12 months (T3). Results: In the autistic sample, negative correlations were observed between self-compassion and subsequent anxiety/depression across all models and timepoints, and these relationships were not reciprocal (i.e., earlier depression and anxiety did not predict future self-compassion). In the non-autistic sample, the findings generally suggested bi-directional relationships between self-compassion and anxiety/depression. In both groups, an indirect pathway between T1 autistic traits and T3 anxiety/depression via T2 self-compassion was confirmed. Conclusion: Considering the high prevalence of anxiety and depression among autistic people, and that self-compassion can be cultivated through practice, these findings suggest that self-compassion could be a useful therapeutic target to promote mental health in the autistic population.