Many countries rely on regional climate model (RCM) projections to quantify the impacts of climate change and to design their adaptation plans accordingly. In several European regions, RCMs project a smaller temperature increase than global climate models (GCMs), which is hypothesised to be due to discrepant representations of topography, cloud processes, or aerosol forcing in RCMs and GCMs. Additionally, RCMs do generally not consider the vegetation response to elevated atmospheric CO2 concentrations; a process which is, however, included in most GCMs. Plants adapt to higher CO2 concentrations by closing their stomata, which can lead to reduced transpiration with concomitant surface warming, in particular, during temperature extremes. Here we show that embedding plant physiological responses to elevated CO2 concentrations in an RCM leads to significantly higher projected extreme temperatures in Europe. Annual maximum temperatures rise additionally by about 0.6 K (0.1 K in southern, 1.2 K in northern Europe) by 2070–2099, explaining about 67% of the stronger annual maximum temperature increase in GCMs compared to RCMs. Missing plant physiological CO2 responses thus strongly contribute to the underestimation of temperature trends in RCMs. The need for robust climate change assessments calls for a comprehensive implementation of this process in RCM land surface schemes.