Abstract Aims Management of patients with atrial fibrillation (AF) and concomitant heart failure (HF) remains complex. The Antwerp score, based on four parameters [QRS >120 ms (2 points), known aetiology (2 points), paroxysmal AF (1 point), severe atrial dilation (1 point)] adequately estimated the probability of left ventricular ejection fraction (LVEF) recovery after AF ablation in a single-centre cohort. The present study aims to externally validate this prediction model in a large European multi-centre cohort. Methods and results A total of 605 patients (61.1 ± 9.4 years, 23.8% females, 79.8% with persistent AF) with HF and impaired LVEF (<50%) undergoing AF ablation in 8 European centres were retrospectively identified. According to the LVEF changes at 12-month echocardiography, 427 (70%) patients fulfilled the ‘2021 Universal Definition of HF’ criteria for LVEF recovery and were defined as ‘responders’. External validation of the score yielded good discrimination and calibration {area under the curve 0.86 [95% confidence interval (CI) 0.82–0.89], P < .001; Hosmer–Lemeshow P = .29}. Patients with a score < 2 had a 93% probability of LVEF recovery as opposed to only 24% in patients with a score > 3. Responders experienced more often positive ventricular remodelling [odds ratio (OR) 8.91, 95% CI 4.45–17.84, P < .001], fewer HF hospitalizations (OR 0.09, 95% CI 0.05–0.18, P < .001) and lower mortality (OR 0.11, 95% CI 0.04–0.31, P < .001). Conclusion In this multi-centre study, a simple four-parameter score predicted LVEF recovery after AF ablation in patients with HF and discriminated clinical outcomes. These findings support the use of the Antwerp score to standardize shared decision-making regarding AF ablation referral in future clinical studies.