BackgroundThis study addresses the nexus of stress, physical exercise, and internet addiction in college students, aiming to unravel their complex interactions. With prior research hinting at correlations, this study seeks deeper insights into the optimal exercise duration and the mediating role of physical activity in the stress-internet addiction relationship. MethodsWe examined 2892 Chinese college students, assessing psychosocial traits via established scales: the Chinese Internet Addiction Scale (CIAS-R), International Physical Activity Questionnaire (IPAQ), and DASS-21 for stress. We adopted a questionnaire survey method to collect data, and through convenience sampling, participants were selected. Analysis involved Structural Equation Model (SEM) and restricted cubic splines (RCS). ResultsAmong participants, 22.0 % self-reported internet addiction symptoms. Notably, a non-linear correlation between moderate (MPA) and vigorous physical activity (VPA) and internet addiction was evident (P < 0.001), while light physical activity (LPA) showed no significant association (P = 0.182). MPA initially reduced internet addiction as activity hours increased, but its efficacy lessened after 0.8 h/day. Conversely, VPA's impact fluctuated, declining from 0 to 0.18 h/day, then stabilizing. SEM analysis highlighted VPA and MPA's partial mediation in the stress-internet addiction relationship. Specifically, the path coefficient from stress to MPA was −0.037 (P < 0.05), and the path coefficient from MPA to internet addiction was −0.056 (P < 0.01). Additionally, the path coefficient from stress to VPA was −0.067 (P < 0.001), and the path coefficient from VPA to internet addiction was −0.102 (P < 0.001). ConclusionThis study unveils the intricate interplay between physical activity, stress, and internet addiction among college students. With 22.0 % reporting addiction symptoms, urgent interventions tailored to promoting healthy physical activity levels are imperative to mitigate internet addiction risks. LimitationsThe cross-sectional nature of our design precludes the establishment of causal relationships between psychological stress and internet addiction.