The task of precise Land Use/Land Cover (LULC) mapping in mountainous areas is a significant challenge due to the complexity of the terrain and spectral confusion. High-resolution global products like Esri’s 10-m LULC maps present valuable opportunities, yet their performance in these contexts remains underexplored. This study aims to fill this gap by evaluating and improving the precision of Esri 10-m LULC datasets (2017 and 2021) in the mountainous areas of Morocco. The results of this study are of practical significance for more accurate land use monitoring. To maximize map coherence, we used stratified random sampling and high-resolution Google Earth imagery to assess classification accuracy, then refined the classification. The results revealed a wide range of classification accuracy for mountainous units (OA ranges between 69.3% and 93.4%), and the misclassification of the “Built Area” class was also found. Particularly in more diverse landscapes, the calibration steps significantly improved accuracy and visual coherence. Analysis of Esri’s 10-m LULC maps further showed substantial LULC changes, including losses of forests, water bodies and expansion of rangelands and built-up areas. In light of the observed limitations, especially in detecting fine-scale anthropogenic features, we propose the integration of artificial intelligence tools—alongside remote sensing data and field verification—as a promising approach to improve the detection and delineation of critical land features, thereby reinforcing the reliability of LULC products in complex mountainous terrains.