In order to test a newly developed algorithm for detecting atrial fibrillation in clinical practice, we carried out parallel recordings using a conventional 24-h electrocardiogram (ECG) monitor and telemonitoring with an external loop recorder. Recordings were made in 24 patients with persistent atrial fibrillation and in another 24 patients with sinus rhythm. Atrial fibrillation was detected immediately in 23 of 24 patients with persistent atrial fibrillation and 20 min after fitting the single-channel loop recorder in the 24th patient (sensitivity 100%). On average, 3.1 false positives (i.e. detection of an episode, including the end or beginning of atrial fibrillation) were transmitted per patient. The sensitivity of the algorithms for automatically detecting bradycardiac and tachycardiac atrial fibrillation was also high. In 12 of 24 patients with sinus rhythm, false-positive tele-ECGs were transmitted. These were caused by supraventricular or ventricular extrasystoles and by sinus arrhythmias or sinoatrial (SA) blocks. The external loop recorder was very effective at detecting paroxysmal atrial fibrillation. Possible indications for the clinical use of this recorder include, in addition to diagnosis, monitoring patients for atrial fibrillation recurrence after cardioversion or catheter ablation.