ObjectiveTo reveal the reciprocal predictive relationship between depressive symptoms (PHQ-9) and internet addiction tendencies (IAT) among first-year university students, and to compare differences across urban-rural backgrounds and only-child status, providing a basis for targeted interventions.MethodsThis study conducted follow-up surveys of 6, 070 freshmen of the 2023 cohort at a university in Jingzhou, Hubei Province, in October 2023 and May 2024. Controlling for gender, household economic status, and family structure, cross-lagged path analysis was used to examine the dynamic association between depression symptoms and internet addiction tendencies.ResultsT1 and T2 PHQ-9 and IAT scores were significantly positively correlated (T1: r=0.453; T2: r=0.485, p<0.01). Nonparametric tests showed significant urban–rural differences for T1 PHQ-9 (Z = 3.069, p=0.002), T2 PHQ-9 (Z = 4.329, p<0.001), T1 IAT (Z = 5.079, p<0.001), and T2 IAT (Z = 5.335, p<0.001); significant only-child differences for T1 PHQ- (Z = 4.173, p<0.001), T2 PHQ-9 (Z = 4.980, p<0.001), T1 IAT (Z = 7.189, p<0.001), and T2 IAT (Z = 6.040, p<0.001). Cross-lagged analyses of the full model indicated bidirectional prediction. In subgroup models, T1 IAT significantly predicted T2 PHQ-9 in all groups: urban only-child (β=0.172, p<0.001), urban non-only-child (β=0.173, p<0.001), rural only-child (β=0.233, p<0.001), and rural non-only-child (β=0.238, p<0.001). T1 PHQ-9 significantly predicted T2 IAT only in urban only-child (β=0.109, p<0.001) and rural non-only-child (β=0.099, p<0.001), but not in urban non-only-child or rural only-child.ConclusionsInternet addiction tendency is significantly correlated with depressive symptoms. At the group level, the baseline level of internet addiction tendency positively predicts subsequent changes in depressive symptoms, and this effect holds true across all groups. The predictive effect of depressive symptoms on subsequent internet addiction tendency is significant only in the urban only-child and rural non-only-child groups in subgroup analyses. However, further multi-group comparisons show that the between-group difference in this path is not statistically significant, so a definitive group moderation effect cannot yet be inferred.