The introduction of the Phase Tensor marked a major breakthrough in the understanding, analysis and treatment of galvanic distortion of the electric field in the Magnetotelluric (MT) method. We build upon a recently formulated impedance tensor decomposition into the known Phase Tensor and an Amplitude Tensor that is shown to be complementary and algebraically independent of the Phase Tensor. This recent decomposition demonstrates that the Amplitude Tensor contains inductive and galvanic information of the subsurface and that the inductive information is physically coupled to the one contained in the Phase Tensor. It is also demonstrated that, through this coupling, galvanic effects can be separated from inductive effects in the Amplitude Tensor. In this work we present an algorithm that employs this last finding to show that the MT galvanic electric distortion tensor can be separated from the inductive Amplitude Tensor and hence, that this distortion can be recovered for any given data up to a single constant usually denoted as galvanic shift or site gain. Firstly, to illustrate distortion effects on the Amplitude Tensor, we manually apply distortion by matrix multiplication to synthetic impedance tensor data. Then, we use the observations of that analysis to define an objective function, which minimises when there is no distortion present in the Amplitude Tensor. Secondly, we describe our algorithm that employs a genetic algorithm to find the optimal distortion tensor needed to correct the Amplitude Tensor, and therewith the impedance tensor. Lastly, we test the performance of the proposed methodology on synthetic data of known distortion, on a large scale (144 sites) synthetic data set of random distortion and on four real data sets taken from the BC87 data set that are reported to contain 3D inductive effects. The real data sets, lit007 /lit008 and lit901 /lit902, demonstrate the utility of the proposed algorithm by revealing geological expected results in the impedance data for the first time. This could not be achieved before by alternative methods due to their inherent assumption of a 2D regional impedance, which is not required in our scheme.