<p>Lithium (Li) is a critical mineral for the global energy transition, yet its occurrence in many regions remains poorly understood due to limited geological data and the complex nature of its distribution. This complexity arises because Li can form via multiple geological processes (genetic models), and each type of deposit is governed by a specific set of geological conditions including host rocks, tectonic settings, and fluid chemistry, which together define the mineral system controlling its formation. Therefore, identifying prospective Li exploration targets with a reasonable level of confidence remains challenging. Supervised machine learning (ML) and artificial intelligence (AI)-based approaches can improve prediction accuracy and reduce uncertainty in mineral prospectivity mapping (MPM). This study applied an advanced GIS- and AI-based toolkit (EIS toolkit) with a number of ML algorithms for producing and comparing MPM for Li-bearing granites and pegmatites in Västernorrland, Sweden. In particular, we used a knowledge-driven fuzzy logic (FL) approach and a data-driven random forest (RF) algorithm, to integrate multisource geoscientific datasets to predict prospective targets for Li mineralization. Input evidence maps were derived from geological, geochemical, geophysical, and structural datasets linked to different components of a conceptual Li mineral system model. Both FL- and RF-derived prospective areas show significant overlap with known Li-bearing pegmatites/granites while also revealing several new target areas for further investigation. The results highlight the complementary strengths of FL and RF methods in integrating diverse datasets and enhancing regional-scale targeting for critical mineral resources. Incorporating forthcoming harmonized and up-to-date bedrock maps, along with expanded lithogeochemical coverage in the entire area, is expected to substantially improve the model accuracy of future MPM and enhance exploration targeting for Li-bearing pegmatites and granites.</p>