The application of a priori information to stabilize the inversion of MT data usually has been done in a purely mathematical form, mainly as smoothness constrains that have little regard to the physical and geological conditions that generate the data. We present a stable solution for the anisotropic magnetotelluric inverse problem through the use of approximate equality constraints as a way of including realistic a priori information in the automatic inversion process. Information on the structures of the earth down to a certain depth can be acquired from well log data or other geophysical methods or geological knowledge. If the values of some of the parameters to be found in the inversion were known exactly, we should simply remove those parameters from the inversion process. However, if there is a degree of confidence in that information, but not certainty, then some freedom should be allowed for those parameters to converge to values that are close, but not exactly equal to those of the constraints. Following that idea, we constrain the solution of the inverse problem by imposing the values of the constraints to the inversion parameters in a least squares sense, therefore the term approximate equality.We use this kind of information to constrain our inversion of MT data and we find that we can stabilize the solution for the anisotropic 1-D problem. We demonstrate the method by inverting synthetic data, contaminated by random noise.