Providing cardiopulmonary resuscitation (CPR) to a patient in cardiac arrest introduces artefacts into the electrocardiogram (ECG), corrupting the diagnosis of the underlying heart rhythm. CPR must therefore be discontinued for reliable shock advice analysis by an automated external defibrillator (AED). Detection of ventricular fibrillation (VF) during CPR would enable CPR to continue during AED rhythm analysis, thereby increasing the likelihood of resuscitation success. This study presents a new adaptive filtering method to clean the ECG. The approach consists of a filter that adapts its characteristics to the spectral content of the signal exclusively using the surface ECG that commercial AEDs capture through standard patches. A set of 200 VF and 25 CPR artefact samples collected from real out-of-hospital interventions were used to test the method. The performance of a shock advice algorithm was evaluated before and after artefact removal. CPR artefacts were added to the ECG signals and four degrees of corruption were tested. Mean sensitivities of 97.83%, 98.27%, 98.32% and 98.02% were achieved, producing sensitivity increases of 28.44%, 49.75%, 59.10% and 64.25%, respectively, sufficient for ECG analysis during CPR. Although satisfactory and encouraging sensitivity values have been obtained, further clinical and experimental investigation is required in order to integrate this type of artefact suppressing algorithm in current AEDs.