Background: A novel approach to improve bystander defibrillation for out-of-hospital cardiac arrests is to dispatch and deliver an automated external defibrillator (AED) directly to the suspected cardiac arrest location by drone. The aim of this study was to investigate how often a drone could deliver an AED before ambulance arrival and to measure the median time benefit achieved by drone deliveries. Methods: In this prospective observational study, five AED-equipped drones were placed within two separate controlled airspaces in Sweden, covering approximately 200 000 inhabitants. Drones were dispatched in addition to standard emergency medical services for suspected out-of-hospital cardiac arrests and flight was autonomous. Alerts concerning children younger than 8 years, trauma, and emergency medical services-witnessed cases were not included. Exclusion criteria were air traffic control non-approval of flight, unfavourable weather conditions, no-delivery zones, and darkness. Data were collected from the dispatch centres, ambulance organisations, Swedish Registry for Cardiopulmonary Resuscitation, and the drone operator. Core outcomes were the percentage of cases for which an AED was delivered by a drone before ambulance arrival, and the median time difference (minutes and seconds) between AED delivery by drone and ambulance arrival. Explorative outcomes were percentage of attached drone-delivered AEDs before ambulance arrival and the percentage of cases defibrillated by a drone-delivered AED when it was used before ambulance arrival. Findings: During the study period (from April 21, 2021 to May 31, 2022), 211 suspected out-of-hospital cardiac arrest alerts occurred, and in 72 (34%) of those a drone was deployed. Among those, an AED was successfully delivered in 58 (81%) cases, and the major reason for non-delivery was cancellation by dispatch centre because the case was not an out-of-hospital cardiac arrest. In cases for which arrival times for both drone and ambulance were available (n=55), AED delivery by drone occurred before ambulance arrival in 37 cases (67%), with a median time benefit of 3 min and 14 s. Among these cases, 18 (49%) were true out-of-hospital cardiac arrests and a drone-delivered AED was attached in six cases (33%). Two (33%) had a shockable first rhythm and were defibrillated by a drone-delivered AED before ambulance arrival, with one person achieving 30-day survival. No adverse events occurred. AED delivery (not landing) was made within 15 m from the patient or building in 91% of the cases. Interpretation: AED-equipped drones dispatched in cases of suspected out-of-hospital cardiac arrests delivered AEDs before ambulance arrival in two thirds of cases, with a clinically relevant median time benefit of more than 3 min. This intervention could potentially decrease time to attachment of an AED, before ambulance arrival. Funding: Swedish Heart Lung Foundation. © 2023 The Author(s).