Background <p>In the digital era, Social Media Addiction has emerged as a prominent issue affecting the mental health of university students. Despite evidence linking physical exercise to lower Social Media Addiction, it remains unknown whether within-person changes in physical exercise actually predict subsequent changes in Social Media Addiction, and whether psychological resilience accounts for this link. Prior work is constrained by cross-sectional designs, the absence of a unified model integrating these three constructs, and reliance on traditional cross-lagged panel models that cannot separate between-person from within-person effects.</p> Objective <p>This study employed a random intercept cross-lagged panel model (RI-CLPM) to examine the longitudinal relationships between physical exercise and Social Media Addiction among university students, and to test the mediating role of psychological resilience.</p> Methods <p>A three-wave longitudinal study was conducted with 632 university students (M_age = 20.29, SD = 0.86) from two universities in China, with data collected in January 2025, July 2025, and January 2026.</p> Results <p>At the between-person level, physical exercise was negatively associated with Social Media Addiction (β = -0.571). At the within-person level, bidirectional relationships were identified: physical exercise negatively predicted subsequent Social Media Addiction (β = -0.199 to -0.138), while Social Media Addiction negatively predicted subsequent physical exercise (β = -0.146 to -0.242). Psychological resilience significantly mediated the effect of physical exercise on Social Media Addiction (β = -0.115), but the reverse pathway was not significant.</p> Conclusion <p>Physical exercise and Social Media Addiction demonstrate a reciprocal relationship at the within-person level, with psychological resilience serving as a significant mediator. Universities should integrate physical exercise into mental health promotion strategies to prevent Social Media Addiction.</p>