Correlation length in atrial fibrillation (AF) can be calculated from intracardiac electrograms (EGMs) and conceptualized as the distance over which electrical signals remain synchronized. Emerging data shows correlation length predicts AF termination and is associated with outcomes after AF ablation. The capacity to estimate correlation length non-invasively would be a major advance for its applicability to clinical care. Association with non-invasive electrocardiographic (ECG) measurements is currently unknown. We sought to determine if correlation length, measured from EGMs at the time of AF ablation, can be predicted non-invasively from the surface ECG in AF. We conducted a prospective, multicentre, observational study, with a published and registered protocol (ACTRN12619001172190p) of participants undergoing clinically indicated AF ablation. Data collected included (1) detailed clinical phenotyping of participants pre-procedurally, (2) intraprocedural mapping of spontaneous or induced AF with a HD Grid catheter, recording 1-minute EGM segments in 6 right atrial and 10 left atrial anatomic locations, sampled at 2000Hz, with simultaneously acquired ECG. Correlation length is a measure of the distance over which EGMs remain synchronized, measured by calculating the cross-correlation of nearby EGMs as a function of distance. Predictors of correlation length were modelled with a multilevel mixed-effects gamma generalised linear regression. The study recruited n=42 participants. Correlation length could be predicted by surface ECG metrics, including ECG dominant frequency (DF) (coefficient -0.0509, 95% CI -0.0881- -0.0136, p=0.007), Hursts exponent (coefficient -11.66, 95% CI -20.90 - -2.42, p=0.013), Kurtosis (coefficient -0.0034, 95%CI -0.0059 - -0.0009, p=0.007), and Shannon entropy (coefficient -0.1701, 95% CI -0.3319- -0.0083, p=0.039) (see figure 1). The directionality of these findings is biologically plausible. ECG features which did not predict the invasive correlation length included ECG standard deviation (p=0.466), ECG skewness (p=0.332), ECG period (p=0.314) and ECG mean (p=0.261). Intracardiac correlation length is associated with surface ECG features. Importantly, ECG features are measurable non-invasively, meaning this approach could be expanded to patients not undergoing AF ablation, or prior to AF ablation. Given correlation length associates with clinical parameters and outcomes, this metric may help to personalize AF therapy in the future.