Aim: The aim was to investigate the clinical performance of two different types of automated external defibrillators (AEDs). Methods: Three investigators reviewed 2938 rhythm analyses performed by AEDs in 240 consecutive patients (median age 72, q1-q3 = 62-83) who had suffered cardiac arrest between January 2011 and March 2015. Two different AEDs were used (AED A n = 105, AED B n = 135) in-hospital (n = 91) and out-of-hospital (n = 149). Results: Among 194 shockable rhythms, 17 (8.8%) were not recognized by AED A, while AED B recognized 100% (n = 135) of shockable episodes (sensitivity 91.2 vs 100%, p < 0.01). In AED A, 8 (47.1%) of these episodes were judged to be algorithm errors while 9 (52.9%) were caused by external artifacts. Among 1039 non-shockable rhythms, AED A recommended shock in 11 (1.0%), while AED B recommended shock in 63 (4.1%) of 1523 episodes (specificity 98.9 vs 95.9, p < 0.001). In AED A, 2 (18.2%) of these episodes were judged to be algorithm errors (AED B, n = 40, 63.5%), while 9 (81.8%) were caused by external artifacts (AED B, n = 23, 36.5%). Conclusions: There were significant differences in sensitivity and specificity between the two different AEDs. A higher sensitivity of AED B was associated with a lower specificity while a higher specificity of AED A was associated with a lower sensitivity. AED manufacturers should work to improve the algorithms. In addition, AED use should always be reviewed with a routine for giving feedback, and medical personnel should be aware of the specific strengths and shortcomings of the device they are using. (C) 2017 Elsevier B.V. All rights reserved.