Long commuting times may constrain opportunities for exercise, but evidence remains limited on whether built-environment associations vary across area-level commuting-time burden. This cross-sectional spatial study analyzed 229,563 June 2024 Keep walking, running and cycling trajectories aggregated to 5802 500 m × 500 m grid cells in central Jinan, China. A within-mode, sample-relative activity index based on GPS-reconstructed distance and mean speed classified trajectories into low-, medium- and high-activity groups. Four theory-defined built-environment composites were examined using PPML models with distance to the urban-core reference, district fixed effects and 391 commuting-location clusters; primary models included 5012 grids. Green/recreation, street/public-transport accessibility and development intensity were positively associated with total and activity-specific trajectory counts in the primary models, whereas daily service provision was not independently associated. Continuous commuting time was not statistically associated with any outcome, and no built-environment dimension showed FDR-adjusted overall moderation. Alternative distance–speed weights did not alter the activity-specific conclusions. However, the green/recreation association was substantially attenuated after adjustment for the origin-based observed platform user base, while the street/transit association was sensitive to trajectory assignment and exposure specification. These findings support cautious interpretation of area-level platform-observed activity associations.