A relatively simple land surface model, the Biosphere‐Atmosphere Transfer Scheme (BATS) and the more complex Community Land Model (CLM3.5) coupled to RegCM4 are used to investigate land‐atmosphere feedback processes during the Indian summer monsoon. Model simulations for 27 years show that the mean and interannual variability of rainfall and surface air temperature are affected significantly due to differences in the formulation of evapotranspiration and hydrological processes between BATS and CLM3.5, prescribed land use, land cover data and changes in net radiation. RegCM4 with CLM3.5 (RCLM) shows a reduction in surface moisture flux and precipitation but an increase in surface air temperature over most parts of India as compared to RegCM4 with BATS (RBAT). In terms of the mean and interannual variability of rainfall over central India, RBAT performs better than RCLM. Evapotranspiration over central India is found to be less (more) sensitive to soil wetness variations in RCLM (RBAT) compared to the multimodel estimate from the Global Soil Wetness Project, that is mainly attributed to the differences in ground evaporation and transpiration. Changes in evaporation efficiency between the models also lead to a reduction in the land‐atmosphere coupling strength for precipitation in RCLM. Furthermore, such changes decrease the convective instability over central and eastern India in RCLM leading to weakened convection, reduced large‐scale moisture convergence and precipitation over land. Observations of soil moisture, surface fluxes, and radiation are needed for better understanding and improvement of coupled land‐atmosphere feedbacks in models during the Indian summer monsoon.