Insufficient physical activity (PA) is a significant global concern among adolescents. The relationship between PA avoidance and loneliness has not been thoroughly investigated, nor have these variables been examined concurrently. The present longitudinal study utilized cross-lagged mediation analysis with sex, age, and weight status controlled to explore the associations between PA avoidance, loneliness, and various levels of PA (including total PA, vigorous PA, moderate PA, and walking) among adolescents. An online survey was administered at three time points (from September 2023 to February 2024), Yielding responses from 1,024 students at the first time point (T1), 972 at the second time point (T2), and 978 at the third time point (T3). A reciprocal interaction between PA avoidance and loneliness was found over time (standardized coefficients [β] = 0.08 to 0.17). More specifically, PA avoidance at T1 negatively predicted both vigorous and moderate PA at T2 (β = -0.11 for both measures). Moreover, loneliness at T1 was negatively associated with walking at T2 (β = -0.10), and loneliness at T2 was negatively associated with vigorous PA at T3 (β = -0.13). Additionally, PA avoidance at T1 was indirectly related to vigorous PA at T3 through the mediation of loneliness at T2 (b = -0.020, SE = 0.010, t = 2.00, p = 0.046). These findings contribute to the existing literature by elucidating the dynamic interplay between PA avoidance, loneliness, and PA levels, thereby showing a novel mechanism through which loneliness may lead to decreased PA. Interventions aimed at enhancing psychological well-being, social skills, or enjoyment of PA may prove beneficial in alleviating PA avoidance and loneliness, ultimately fostering a more active lifestyle among adolescents.