Purpose: Physical activity (PA) has been linked to better cognitive ability (CA), but the bidirectionality of this association and the role of activity intensity remain unclear. We aimed to examine bidirectional longitudinal associations between PA and CA across three national cohorts. Methods: We used harmonized data from three population-based aging studies: the English Longitudinal Study of Ageing (ELSA), the US Health and Retirement Study (HRS), and the China Health and Retirement Longitudinal Study (CHARLS). Participants aged ≥50 years with three waves of data were included (n = 15,120). PA was categorized by intensity (light, moderate, vigorous) and self-reported at baseline and follow-up. CA was assessed via composite scores of memory, orientation, and numeracy. We applied linear mixed-effects models to estimate longitudinal bidirectional associations between PA (PA intensity) and CA, adjusting for sociodemographic and health-related covariates. Results: In ELSA cohort, we observed modest but statistically significant bidirectional associations between PA and CA. In CHARLS cohort, the CA-PA association attenuated to non-significance after adjusting for chronic conditions, while the HRS cohort showed similar patterns to CHARLS. Notably, Moderate-intensity PA showed the most consistent longitudinal associations with subsequent cognitive outcomes across all cohorts (β = 0.024, p = 0.001). Evidence for reverse association was also observed across all cohorts (β = 0.016, p = 0.017). No systematic moderation by gender or baseline age was detected. Results were broadly comparable across cohorts, despite differences in measurement and cultural context. Conclusions: Regular moderate-intensity PA may support cognitive health in later life, and cognitive function may in turn influence ongoing engagement in MPA. These findings underscore the value of reciprocal benefits of moderate-intensity PA as cognitive health strategies in aging populations.