This study investigates nonlinear associations between built environment (BE) attributes and commuting mode share within rail and bus rapid transit (BRT) catchment areas. Data from approximately 2,790 fixed guideway transit station areas across 34 metropolitan statistical areas in the United States were analyzed using a random forest approach. Results show nonlinear associations of the BE with mode share in transit catchment areas with substantial differences between rail and BRT catchment areas. Rail catchment areas exhibit greater sensitivity to the BE to reduce car dependency as compared to BRT stations. Moreover, polycentricity and population density at regional level are effective at reducing car dependency. We suggest that policymakers consider BE threshold effects and station types when adjusting land-use policies, e.g., in transit-oriented development. Additionally, BRT systems could be a useful alternative to rail in sprawling areas that do not have the compactness needed to support rail.