The goaf and subsidence areas formed ue to the mining of subsurface coal seams can cause damage to surrounding ecological environments.At present, the detection effects of double-layer goafs in coal mines, especially the second-layer goafs, are unsatisfactory. In response to this challenge, this study delineated goafs using apparent resistivity and impedance phase derived from the data acquired by an efficient controllable source audio-frequency magnetotelluric instrument. In the case of a shallow water-bearing goaf with low resistivity, the apparent resistivity displays shadow effects, leading to an extended abnormal range of the upper target, which is unfavorable to the identification of the lower high-resistivity goaf. In contrast, the impedance phase, exhibiting minor shadow and static effects, shows a significant response to the lower goaf. As indicated by the theoretical model testing results, the combination of apparent resistivity and impedance phase can effectively determine shallow water-bearing goafs and deep unfilled high-resistivity goafs. This combination method was employed to interpret the double-layer goaf in the Shenfu mining area of the Jurassic coal field in northern Shaanxi, achieving satisfactory results through the mutual verification of the two parameters. Engineering verification results indicate that this method demonstrates reliable inference and expected exploration effects. Overall, this method provides a new approach for CSAMT-based inference and interpretation in the exploration of double-layer goafs in coal mines, thus holding critical technical promotion and reference significance.