Quantification of groundwater recharge is highly important for understanding freshwater systems and sustainable management of both groundwater and surface water bodies, but it is very uncertain. To improve the global‐scale simulation of diffuse groundwater recharge (GWR), we developed, for the first time in global hydrological modeling, an approach for explicitly simulating GWR in karst areas, based on the World Karst Aquifer Map and ground‐based estimates of karst GWR. Ground‐based estimates of both karst and non‐karst GWR from 5,600 locations were aggregated to produce estimates for 816 globally distributed 0.5° grid cells, including 64 representing karst GWR. Applying WaterGAP2.2e, we found that karst GWR is best approximated by total runoff from land (without soil overflow and urban runoff). The GWR algorithm for non‐karst GWR was tuned against ground‐based estimates for 422 non‐karst grid cells outside of Australia. While simulated GWR tends to underestimate GWR in non‐karst areas with mean annual ground‐based GWR > 200 mm/yr, it overestimates national GWR in six out of seven European countries. With an increase in the coefficient for the discharge from groundwater to surface water bodies, the updated GWR algorithm results in an improved fit of simulated streamflow to observations, including low flows. Global mean annual GWR and thus renewable groundwater resources (without Greenland and Antarctica) are estimated to be about one‐half of the total renewable water resources, or 21,000 km3/yr, which is over 50% higher than the standard model estimate. 20% of global GWR is generated in karst areas, which cover about 11% of the world's land surface.