Groundwater is an important source of domestic and agricultural water supply in northern Uganda, where surface-water availability is highly seasonal. However, rainfall variability and land-use and land-cover (LULC) change strongly influence groundwater recharge and occurrence. This study assessed the spatio-temporal dynamics of groundwater potential zones (GWPZs) in Agago District between 1992 and 2020 and projected their spatial distribution for 2030 and 2040. We applied an integrated GIS–Remote Sensing–Analytic Hierarchy Process (AHP) framework using seven groundwater-controlling factors: rainfall, geology, land use and land cover, slope, drainage density, lineament density, and soil. Historical climate influence was represented using reconstructed and spatially analyzed rainfall data for 1992–2020. Results indicate that rainfall received the highest AHP weight (37.15%), followed by geology (24.15%) and LULC (16.02%); the pairwise-comparison matrix produced a consistency ratio of 0.03, indicating acceptable weighting consistency. Similarly, the proportion of good-potential areas decreased from 41.14% in 2000 to 27.97% in 2020, while fair-potential areas increased from 58.86% to 72.03%. Additionally, independent validation using 41 borehole-yield records produced ROC–AUC values of 0.965 for the training subset and 0.941 for the validation subset. Future GWPZ patterns were projected using the CA–ANN/MOLUSCE framework based on historical GWPZ transitions and spatial driver variables; no CMIP6 GCM/RCM climate projections or SSP/RCP scenarios were incorporated. The 2020 hindcast achieved 96.51% overall correctness and a kappa coefficient of 0.9246, representing agreement between simulated and observed GWPZ patterns rather than independent borehole-based validation. The findings indicate a contraction of good-potential zones under the observed rainfall and land-use trajectory, highlighting the priority areas for groundwater management, recharge protection, and climate adaptation in northern Uganda.