[Purposes] This paper aims to investigate the impact of built environment on online car?hailing travel demand. [Methods] Taking Chengdu as an example, the research scope is divided into eight spatial units by using different scales of regular grids and traffic analysis areas. For each spatial unit, a multi?scale geographical weighted regression model is used for regression analysis to determine the optimal spatial analysis unit. According to the optimal bandwidth, the built environment influencing factors are divided into local, regional and global variables, and this research analyzes the spatial heterogeneity of the influence degree of built environment variables and their scale difference; and the relationship between the built environment and the demand for online car?hailing is explored. [Findings] The 1?100 m grid is the best spatial unit. Under this spatial unit, the determination coefficients R2 of the regression model of car?hailing demand for the morning peak hours and the evening peak hours are 94.5% and 96.7%, respectively. Moreover, the spatial scales of different built environment variables are different, and the local coefficients have spatial heterogeneity and varies significantly for the morning peak hours and the evening peak hours. [Conclusions] The best spatial analysis unit can provide a reference for the traffic zoning of online car?hailing demand forecasting. When adjusting passenger flow demand for different locations, the results of spatial heterogeneity could provide reference for formulating more reasonable built environment renewal strategies.