We introduce a new seismic-electromagnetic (EM) projection (SEMP) attribute to improve the characterization of pay sands by integrating seismic P impedance and S impedance with electrical resistivity measurements leveraging on the high sensitivity of resistivity to fluid saturation and the capability of seismic data to delineate thin beds. Porosity is predicted using an extended elastic impedance (EEI) approach to reduce the risk of encountering SEMP anomalies related to low porosity. First, we determine the seismic-EM integration concept using well logs from a well that encountered hydrocarbons in a deepwater fold belt in offshore Borneo. The SEMP logs supported by the EEI log successfully delineate the pay zones in the well. Second, we extend the approach in three dimensions and evaluate its effectiveness using large-size seismic controlled-source EM (CSEM) and magnetotelluric (MT) data from the same area. The seismic data are inverted to produce P- and S-impedance volumes. EEI volume calibrated to total porosity also is produced. We then perform 3D anisotropic cross-gradient joint CSEM and MT inversion of a large-size survey comprising 647 receivers with receiver spacing of 1.5–3 km at prospect to regional scale. The SEMP volumes computed using the seismic and CSEM-MT inversion results are then interpreted with the EEI volume to identify possible geobodies of pay sand. Resistive geobodies with low porosity are assigned higher risks because they are likely due to compaction rather than hydrocarbon presence, whereas geobodies with high porosity are assigned lower risks and identified as potential sweet spots. The geobodies need to be corroborated with independent geologic studies to mature them into exploration prospects and assess their associated risk and potential. Although a vertical transverse isotropy approximation is used in our CSEM-MT study, the SEMP concept also is applicable to cases considering tilted transverse isotropy or arbitrary anisotropy where available and operational for such large-size exploration data.