Abstract: As shallow and easily identifiable mineral resources become increasingly depleted, mineral exploration has gradually shifted toward deep, concealed, and covered areas. Traditional exploration methods now face practical bottlenecks, including low data-processing efficiency, difficulty uncovering hidden mineralization patterns, and insufficient integration of multi-source information. With its strong non-linear fitting, multimodal data mining, and intelligent reasoning capabilities, artificial intelligence (AI) is increasingly becoming the core driver of the transformation and upgrading of geological prospecting. This paper systematically reviews the research background of AI-empowered mineral prospecting and categorizes the application status of technologies—such as machine learning, deep learning, knowledge graphs, and retrieval-augmented generation (RAG)—across key workflow stages, including data preprocessing, anomaly identification, metallogenic prediction, target area optimization, and intelligent interpretation. It also analyzes critical issues facing the industry, such as difficulties fusing multi-source heterogeneous data, weak model interpretability, insufficient domain adaptability, and high thresholds for engineering deployment. Finally, in light of trends in geological big data and large models, this paper proposes future directions for technical optimization, framework development, and industrial integration. This study aims to provide a scientific reference for the deep integration of AI and mineral exploration, thereby supporting the implementation of the new round of strategic breakthroughs in mineral prospecting.