<strong>Background:</strong> The COVID-19 pandemic disrupted mobility globally, but its mental health implications remain difficult to characterize because most studies relied on lockdown status, population-level mobility indicators, or self-reported mobility. These approaches may miss individual differences in actual movement patterns and cannot fully examine bidirectional relationships between mobility and mental health. Individual-level smartphone geolocation data may provide a more objective and temporally aligned measure of mobility during periods of societal disruption. <strong>Objective:</strong> This study aimed to use individual-level Google location history (GLH) data and population-level Google community mobility reports (GCMRs) to examine concurrent and longitudinal relationships between pandemic-era mobility patterns and mental health symptoms in Hong Kong. <strong>Methods:</strong> This study analyzed data from the CU-COVID19 cohort study, an online longitudinal survey study of the psychological impact of the pandemic in Hong Kong. Mental health symptoms over the previous 14 days were assessed at baseline, 6 months, and 12 months using the 9-item Patient Health Questionnaire, the 7-item Generalized Anxiety Disorder scale, and the 4-item PTSD Checklist for DSM-5. Participants provided retrospective GLH data reflecting their mobility during the corresponding 14-day survey periods. The analytic sample included 145 participants with baseline GLH data, of whom 110 had 6-month follow-up data and 49 had data available at all 3 assessment waves. GLH data were used to derive mobility factors representing journey diversity, immobility, and remoteness. Population-level mobility during the same 14-day periods was measured using Hong Kong GCMR residential stay data. Concurrent mediation models examined whether individual mobility mediated associations between population-level residential stay and mental health symptoms. Longitudinal models examined bidirectional associations between changes in individual mobility and mental health across 6-month intervals. <strong>Results:</strong> Population-level residential stay was not directly associated with mental health. In concurrent mediation models, higher population-level residential stay was associated with lower individual journey diversity (β=–0.36; <i>P</i><.001), and lower journey diversity was associated with higher depression (β=–0.29; <i>P</i>=.02) and posttraumatic stress disorder (PTSD) (β=–0.35; <i>P</i>=.002). Bootstrapped indirect effects suggested mediation through journey diversity for depressive symptoms (β=0.11, 95% CI 0.02-0.25) and PTSD symptoms (β=0.13, 95% CI 0.05-0.27), although the depression-related indirect effect became less robust after adjustment for local and individual COVID-19 infection indicators. Longitudinally, higher baseline depressive symptoms predicted subsequent reductions in journey diversity (β=–0.15; <i>P</i>=.02), and reductions in journey diversity predicted higher subsequent depressive symptoms (β=–0.43; <i>P</i>=.008). <strong>Conclusions:</strong> Individual-level mobility patterns, particularly lower journey diversity, showed more consistent associations with mental health symptoms than population-level residential stay. Findings suggest bidirectional relationships between mobility and mental health and demonstrate the potential of smartphone geolocation data for digital phenotyping. However, the modest and self-selected sample, limited GCMR availability, and observational design require cautious interpretation.
