The electromagnetic (EM) coupling effect in induced polarization (IP) data is an important problem. In many works it has been computed only considering homogeneous or layered earth models with discretely uniform conductivity. In this study, an algorithm has been developed to compute the EM coupling effect in IP data measured on the earth, whose conductivity varies (increases or decreases) exponentially with depth. The EM coupling effects for Percent Frequency Effect (PFE) and phase data are computed for a dipole-dipole array with different separations, however the method can be applied to any electrode array. The results obtained for the cases of increasing and decreasing conductivity as a function of depth indicate that the EM coupling effect strongly depends on the subsurface resistivity and the dipole length. Here an "exponential" earth model is considered to remove EM coupling from the IP data in frequency and phase domain. For this purpose, first, the region of pseudo-section is divided into segments, and within each segment a typical average apparent resistivity (rho (a)) curve is constructed. An exponential conductivity model is fitted to average rho (a) data. The conductivity model is then used to compute EM responses. Next the data are corrected for the EM coupling contribution. This decoupling process is applied to field data from a galenite-pyrite mineralization area at the Dolluk site, in western Turkey. The results from the decoupling method developed here are compared with other techniques.