<p>Lithological mapping is a fundamental process to understand the distribution, composition, and characteristics of surface and subsurface rocks, offering essential insights into various fields such as geological surveys, mineral exploration, geological hazards, ecology, and hydrology. Traditionally, lithological mapping involves extensive fieldwork that relies on expert knowledge and subjective judgment. Recent advances have revolutionized lithological mapping ranging from fieldwork to the latest artificial intelligence technologies in conjunction with multi-source geospatial data, enabling large-scale and quantitative identification with high resolution and accuracy. This paper provides a comprehensive review of the principles, data sources, state-of-the-art methodologies, applications, and limitations of lithological mapping, and builds three workflows based on geochemical, geophysical, and remote sensing data. Additionally, this paper further explores potential solutions to these challenges, emphasizing the importance of deep learning algorithms and technological innovations in multi-source data fusion, cloud computing, large foundation models, transfer learning, lightweight networks, model interpretability, and uncertainty quantification to enhance the reliability and performance of lithological mapping. In summary, this review synthesizes the current developments and challenges in lithological mapping, and highlights the growing significance of artificial intelligence technologies, not only for improving the precision and efficiency of lithological mapping, but also for transforming geoscientific applications more broadly.</p>