<p>Inversion of controlled source electromagnetic (CSEM) data estimates a conductivity model for the subsurface and thus can directly indicate the occurrence of hydrocarbons. However, the diffusive regime of low-frequency electromagnetic fields in the subsurface compromises the resolution of the estimates. To mitigate this problem, we propose a modified steepest descent inversion strategy for CSEM data (which basically depends on the gradient of the objective function), using an image-guided gradient that carries out reliable structural information from a model of the target area. As a result, at each iteration, the estimates are updated to attain a better fit of the CSEM data with a conductivity distribution compatible with the available structural model. This structural model can be constructed, for instance, from depth-migrated seismic images of the target area. Through selected numerical experiments, we show that the proposed image-guided inversion approach can significantly improve the accuracy and resolution of resistivity estimates compared to the standard steepest descent inversion.</p>