Vagus nerve stimulation (VNS) therapy is associated with laryngeal muscle activation and induces voice modifications, well-known side effects of the therapy resulting from co-activation of the recurrent laryngeal nerve. In this study, we describe the non-invasive transcutaneous recording of laryngeal motor evoked potentials (LMEPs), which could serve as a biomarker of effective nerve activation and individual titration in patients with drug-resistant epilepsy. We recruited drug-resistant epileptic patients treated for at least 6 months with a VNS. Trains of 600-1200 VNS pulses were delivered with increasing current outputs. We placed six skin electrodes on the ventral surface of the neck, in order to record LMEPs whenever the laryngeal muscular threshold was reached. We studied the internal consistency and the variability of LMEP recordings, and compared different methods for amplitude calculation. Recruitment curves were built based on the stimulus-response relationship. We also determined the electrical axis of the LMEPs dipole in order to define the optimal electrode placement for LMEPs recording in a clinical setting. LMEPs were successfully recorded in 11/11 patients. The LMEPs threshold ranged from 0.25 to 1 mA (median 0.50 mA), and onset latency was between 5.37 and 8.77 ms. The signal-to-noise ratio was outstanding in 10/11 patients. In these cases, excellent reliability (Intraclass correlation coefficient, ICC > 0.90 across three different amplitude measurements) was achieved with 10 sample averages. Moreover, our recordings showed very good internal consistency (Cronbach's alpha > 0.95 for 10 epochs). Area-under-the-curve and peak-to-peak measurement proved to be complementary methods for amplitude calculation. Finally, we determined that an optimal derivation requires only two recording electrodes, aligned on a horizontal axis around the laryngeal prominence. In conclusion, we describe here an optimal methodology for the recording of VNS-induced motor evoked responses from the larynx. Although further clinical validation is still necessary, LMEPs might be useful as a non-invasive marker of effective nerve activation, and as an aid for the clinician to perform a more rational titration of VNS parameters.