Feedback-control has been proved to be advantageous in various technical fields and is likely to increase the performance of electrical neural interface devices. The control algorithms in such a device will rely on metrics of neural activity, thereby necessitating their differentiation from artifacts caused by electrical stimulation. We demonstrate an efficient algorithm for determining the relationship between the electrical stimulus current waveform and the recorded artifact potential, or transfer function. This facilitates online stimulus artifact subtraction and concurrent neural recordings during electrical stimulation. Furthermore, we demonstrate significant changes in this transfer function, in vivo, that occur on time scales of hours and are indicative of changes in the electrical properties of neural tissue. Tracking these variations is paramount for the successful implementation of a feedback-enabled neural control system.