A detailed geochemical study of fluids from representative quartz-sealed faults hosting late Hercynian gold concentrations shows that fluids percolating the mineralised faults had two main distinct reservoirs: one was a quite shallow and the other rather deep-seated. Both fluids have lost a great part of their original geochemical signature through interactions with host metamorphic formations. Early fluids, present during the primary sealing of the faults by quartz, are considered to have effectively equilibrated with the metamorphic pile and then predominantly flowed upwards along the faults. They are characterised by CH4/CO2/H2O ratios rather typical of fluids equilibrated with graphite, and moderate to medium chlorinities with a high Br/Cl ratio. The striking feature of the gold-bearing quartz is that gold is not synchronous within any quartz deposition, and appears located in late microfractures and associated with Pb–Bi–Sb sulphosalts and sulphides. These late stages are characterised by fluids whose salinities decrease to very low values indicating their progressive dilution by waters of more surficial origin in the fault system. The long-lived activity of the fault favoured the connection between two distinct fluid reservoirs at a critical time during the basement uplift. The fluids evolved through two main driving mechanisms which were responsible for the Au deposition: (i) decrease in temperature accompanying decompression from supra-lithostatic to hydrostatic conditions which yielded, in some instances, volatile unmixing in the faulted systems, (ii) mixing of the resulting fluids with waters entering the hydrological systems from shallower reservoirs. In addition to dilution and fluid mixing which are favourable factors for decreasing the gold solubility, the presence of microfractured sulphides could have enhanced gold precipitation through electrochemical processes.