222Rn anomalies that occurred at the time of the 19 September 1985 M8.1 Michoacan, Mexico, earthquake in fluids from two wells of Los Azufres, geothermal field were studied considering (i) the changes in well-bottom thermodynamic conditions (pressure, enthalpy and steam/water ratio) and (ii) the discharged mass flow rates. In the studied steam well the anomalously high 222Rn concentration (as compared with background data) was observed on 11 September and was due to an important compressive process, which formed two-phase fluids at reservoir, where usually single steam occurred. This caused the increase in 222Rn, as it remained in the steam. The compressive process was partially induced by changes in the production conditions at wellhead (because a well test was being developed) when sampling, but it was also seismic-induced, as similar 222Rn anomalies were observed in other steam wells for which stable conditions at sampling were maintained. Another case is given for a two-phase well where anomalously low 222Rn values (compared to background data) were recorded in September–October 1985. As stable conditions were maintained during sampling (1983–1986), 222Rn data were correlated to ?twb> the total mass flow rates produced by the well. Results suggested that the well produced two types of liquid–steam mixtures with a common end-member regarding the liquid, but implying two ‘end-member’ fluids with different 222Rn contents. The most 222Rn enriched end-member was regarded as the original reservoir steam while that with relatively lower 222Rn concentration was stated as likely constituted by a mixture of reservoir and reinjection fluids. From the linear correlations, 222Rn contents for both end-members were estimated while the 222Rn composition of the reservoir liquid feeding the studied well was estimated to be negligible compared to steam from the intersection of both curves. As one anomalously low 222Rn value was measured for a sample collected on the 12 September 1985, it seemed that a seismic-induced process increased the reservoir permeability just before the earthquake allowing the liquid to reach the production zones of some two-phase wells. This interpretation was supported by the behaviour of gas species (CO2 and N2) and δD in the two-phase well studied and injection mass flow rates in one probably-related reinjection well.