Abstract Aims Brugada syndrome (BrS) represents a major cause of sudden cardiac death in young individuals. The risk stratification to forecast future life-threatening events is still controversial. Non-invasive assessment of late potentials (LPs) has been proposed as a risk stratification tool. However, their nature in BrS is still undetermined. The purpose of this study is to assess the electrophysiological determinants of non-invasive LPs. Methods and Results Two hundred and fifty consecutive patients with (Group 1, n = 96) and without (Group 2, n = 154) BrS-related symptoms were prospectively enrolled in the registry. Signal-averaged electrocardiogram (SAECG) was performed in all subjects before undergoing epicardial mapping. Group 1 patients exhibited larger arrhythmogenic substrates (AS; 5.8 ± 2.8 vs. 2.6 ± 2.1 cm2, P < 0.001) with more delayed potentials (220.4 ± 46.0 vs. 186.7 ± 42.3 ms, P < 0.001). Late potentials were present in 82/96 (85.4%) Group 1 and in 31/154 (20.1%) Group 2 individuals (P < 0.001). Patients exhibiting LPs had more frequently a spontaneous Type 1 pattern (30.1% vs. 10.9%, P < 0.001), SCN5A mutation (34.5% vs. 21.2%, P = 0.02), and exhibited a larger AS with longer potentials (5.8 ± 2.7 vs. 2.2 ± 1.7 cm2; 231.2 ± 37.3 vs. 213.8 ± 39.0 ms; P < 0.001, respectively). Arrhythmogenic substrate dimension was the strongest predictor of the presence of LPs (odds ratio 1.9; P < 0.001). An AS area of at least 3.5 cm2 identified patients with LPs (area under the curve 0.88, 95% confidence interval 0.843–0.931; P < 0.001) with a sensitivity of 86%, specificity 88%, positive predictive value 85%, and negative predictive value 89%. Conclusion The results of this study support the role of the epicardial AS as an electrophysiological determinant of non-invasive LPs, which may serve as a tool in the non-invasive assessment of the BrS substrate, as SAECG-LPs could be considered an expression of the abnormal epicardial electrical activity. ClinicalTrials.gov number (NCT02641431; NCT03106701).