Hydrothermal mineral systems are formed by the transport of metals from large source areas through convective fluid flow, subsequent leading to deposition of these metals at specific sites. The fluid pathways are crucial for connecting mineral sources with favorable zones of mineral deposition. However, due to the complexity of fluid flow and limitations in sampling cost, assay cost, and expert experience, inferring fluid pathways poses a significant challenge. In this paper, we leverage the continuous and extensive characteristics of exploration data to identify fluid pathways in hydrothermal deposits, uncovering the hidden patterns from their mineralization footprints and favorable structural features within the data. By modeling the fluid flow as a Markov process, we tailor a hidden Markov model (HMM) to identify fluid pathways using observations of mineralization and structural features. Specifically, we identify the latent geometry of fluid pathways by maximizing their posterior probability as represented by the HMM. We then represent the identified fluid pathways as two quantitative and mappable exploration criteria—trajectory length and pathway flux—which serve as predictor variables in 3D mineral prospectivity mapping. Our method is applied to the Xiadian orogenic gold deposit in the Jiaodong Peninsula, China. The results suggest that the formation of Xiadian deposit is attributed to a series of fluid trajectories originating from two injection points. By using the exploration criteria derived from the identified fluid pathways, we significantly enhance the accuracy and efficacy of mineral prospectivity mapping, demonstrating the proposed HMM as an effective artificial intelligence tool for mineral exploration targeting.