Background Obstructive sleep apnea (OSA) is strongly associated with atrial fibrillation (AF). Long-term ECG monitoring with implantable loop recorders facilitates the identification of undiagnosed AF in 20% of severe OSA cases. However, ambulatory ECG (AECG) monitoring is less resource intensive, and various parameters have been shown to predict AF. The aim of this study was to assess the efficacy of such AECG-based AF predictors in identifying patients with severe OSA most at risk. Methods Prospective observational study including patients with severe OSA and no history of AF. Patients had two 24-h AECG recordings, and if no AF was detected, implanted with a loop recorder (maximum 3 years). Results Of 25 patients implanted, AF ≥ 10 s was detected in 5 patients. None of the parameters from the AECG recordings were significantly different between patients who did and did not develop AF. Conclusions AECG-based parameters were not effective for the prediction of AF in this severe OSA cohort.