New dimensions for geological mapping and mineral prediction have been opened by hyperspectral remote sensing endowed with nanometer-level spectral resolution, and the intelligent transformation of mineral resource exploration has been driven by its deep integration with artificial intelligence (AI). Research advances in this field were systematically reviewed, including the advantages of hyperspectral data for geological mapping and mineral prediction, state-of-the-art machine-learning and deep-learning approaches with comparative analyses, application scenarios of different training strategies, and advanced methods for other typical hyperspectral tasks. Existing challenges were summarized and future prospects for AI-driven hyperspectral geological mapping and mineral resource prediction were outlined. From three perspectives of innovative methodologies, learning strategies, and application paradigms, a theoretical framework for intelligent mineral-resource exploration was constructed, and directional guidance was provided for the interdisciplinary development of hyperspectral remote sensing and next-generation AI technologies.