Background: An atrioventricular defibrillator system with a floating atrial dipole (VDD ICD) can provide atrial sensing by a single lead. Our aim was to compare the arrhythmia detection efficacy of VDD ICDs with conventional single- (VVI) and dual-chamber (DDD) defibrillators. Methods: Data from consecutive patients undergoing ICD implantation were retrospectively analyzed. The primary endpoint was the incidence of device-detected, new-onset atrial arrhythmias, while secondary endpoints were sensing parameters, complication rates, incidence of appropriate/inappropriate ICD therapy, arrhythmic/heart failure-related hospitalizations, and all-cause mortality. Results: A total of 256 patients (mean age 64 ± 12 years, male 75%, primary prophylaxis 28%, mean follow-up 3.7 ± 2.4 years) were included (VVI: 93, VDD: 94, DDD: 69). Atrial arrhythmia episodes were detected more frequently by VDD systems compared to VVI ICDs (aHR 7.087; 95% CI 2.371–21.183; p < 0.001), and at a rate similar to that of DDD ICDs (aHR 1.781; 95% CI 0.737–4.301; p = 0.200). The rate of inappropriate shocks was not different among the three ICD systems. Conclusion: VDD devices revealed an advantage in atrial arrhythmia detection compared to VVI ICDs and were non-inferior to DDD systems. Their main indication may be closer monitoring in high-risk patients with atrial arrhythmias to help therapy optimization and not the improvement of tachycardia discrimination.