The Iberian Peninsula is a water-scarce region that is increasingly reliant on groundwater. Climate change is expected to exacerbate this situation due to projected irregular precipitation patterns and frequent droughts. Here, we utilised convolutional neural networks (CNNs) to assess the direct effect of climate change on groundwater levels, using monthly meteorological data and historical groundwater levels from 3829 wells. We considered temperature and antecedent cumulative precipitation over 3, 6, 12, 18, 24, and 36 months to account for the recharge time lag between precipitation and groundwater level changes. Based on CNNs performance, 92 location-specific models were retained for further analysis, representing wells spatially distributed throughout the peninsula. The CNNs were used to assess the influence of climate change on future groundwater levels, considering an ensemble of eight combinations of general and regional climate models under the RCP4.5 and RCP8.5 scenarios. Under RCP4.5, an average annual temperature increase of 1.7 °C and a 5.2 % decrease in annual precipitation will result in approximately 15 % of wells experiencing >
1-m decline between the reference period [1986-2005] and the long-term period [2080-2100]. Under RCP8.5, with a 3.8 °C increase in temperature and a 20.2 % decrease in annual precipitation between the same time periods, 40 % of wells are expected to experience a water level drop of >
1 m. Notably, for 72 % of the wells, temperature is the main driver, implying that evaporation has a greater impact on groundwater levels. Effective management strategies should be implemented to limit overexploitation of groundwater reserves and improve resilience to future climate changes.