Background Compound motor action potential (CMAP) monitoring is a common method used to prevent right phrenic nerve palsy during cryoballoon ablation for atrial fibrillation. Objective We compared recordings simultaneously obtained with surface and hepatic electrodes. Methods We included 114 consecutive patients (mean age 61.7 ± 10.9 years) admitted to our department for cryoballoon ablation. CMAP was monitored simultaneously with a hepatic catheter and a modified lead I ECG, whilst right phrenic nerve was paced before (stage 1) and during (stage 2) the right-sided freezes. If phrenic threat was detected with hepatic recordings (CMAP amplitude drop >30%) the application was discontinued with forced deflation. Results The ratio of CMAP/QRS was 4.63 (2.67–9.46) for hepatic and 0.76 (0.55–1.14) for surface ( p < 0.0001). Signal coefficients of variation during stage 1 were 3.92% (2.48–6.74) and 4.10% (2.85–5.96) ( p = 0.2177), respectively. Uninterpretable signals were more frequent on surface (median 10 vs. 0; p < 0.0001). For the 14 phrenic threats, the CMAP amplitude dropped by 35.61 ± 8.27% on hepatic signal and by 33.42 ± 11.58% concomitantly on surface ( p = 0.5417). Our main limitation was to achieve to obtain stable phrenic capture (57%). CMAP monitoring was not reliable because of pacing instability in 15 patients (13.16%). A palsy occurred in 4 patients (3.51%) because cryoapplication was halted too late. Conclusion Both methods are feasible with the same signal stability and amplitude drop precocity during phrenic threats. Clarity and legibility are significantly better with hepatic recording (sharper signals, less far-field QRS). The two main limitations were pacing instability and delay between 30% CMAP decrease and cryoapplication discontinuation.