Urban greenspace (UGS) is a crucial nature-based solution for mitigating increasing human exposure to extreme heat, but its long-term potential has been poorly quantified. We used high spatial-temporal resolution data sets of urban land cover and population grid in combination with an urban climate model, machine learning, and land use simulation model to assess the impact of UGS on population exposure to extreme (high-heat exposure, HHE) and its potential spatial optimization strategies. Results showed that the UGS and HHE have a strong spatiotemporal dynamic coupling in 21st century Chinese cities. Moreover, UGS shrinkage increased the HHE by 0.58-1.15 °C, while UGS expansion mitigated it by 0.72-1.26 °C, both stronger in the SSP3-7.0 and SSP5-8.5 scenarios. Different from common impressions, spatial relationships, rather than quantities of UGS, are more influential (1.3-1.8 times) on HHE. Our solutions suggest that simply enhancing the spatial dynamic connectivity between patches can mitigate HHE by 9.1-21.1%, especially for the eastern and central cities. Our results provide an example of how to improve climate adaptation in urban ecological space designs and strongly promote research on optimal spatial patterns for future robust urban heat mitigation.