A diverse range of organisms utilize neurotoxins that target specific ion channels and modulate their activity. Typically, toxins are clustered into several multigene families, providing an organism with the upper hand in the never-ending predator-prey arms race. Several gene families, including those encoding certain neurotoxins, have been subject to diversifying selection forces, resulting in rapid gene evolution. Here, we sought a spatial pattern in the distribution of both diversifying and purifying selection forces common to neurotoxin gene families. Utilizing the Mechanistic Empirical Combination model, we analyzed various toxin families from different phyla affecting various receptors and relying on diverse modes of action. By this approach, we were able to detect clear correlations between the pharmacological surface of a toxin and rapidly evolving domains, rich in positively selected residues. On the other hand, patches of negatively selected residues were restricted to the non-toxic face of the molecule and most likely help in stabilizing the tertiary structure of the toxin. We thus propose a mutual evolutionary strategy of venomous animals in which adaptive molecular evolution is directed towards the toxin active surface. Furthermore, we propose that the binding domains of unstudied toxins could be readily predicted using evolutionary considerations.