(1) Background: The exact anatomic localization of the accessory pathway (AP) in patients with Wolff–Parkinson–White (WPW) syndrome still relies on an invasive electrophysiologic study, which has its own inherent risks. Determining the AP localization using a 12-lead ECG circumvents this risk but is of limited diagnostic accuracy. We developed and validated an artificial intelligence-based algorithm (location of accessory pathway artificial intelligence (locAP AI)) using a neural network to identify the AP location in WPW syndrome patients based on the delta-wave polarity in the 12-lead ECG. (2) Methods: The study included 357 consecutive WPW syndrome patients who underwent successful catheter ablation at our institution. Delta-wave polarity was assessed by four independent electrophysiologists, unaware of the site of successful catheter ablation. LocAP AI was trained and internally validated in 357 patients to identify the correct AP location among 14 possible locations. The AP location was also determined using three established tree-based, ECG-based algorithms (Arruda, Milstein, and Fitzpatrick), which provide limited resolutions of 10, 5, and 8 AP locations, respectively. (3) Results: LocAP AI identified the correct AP location with an accuracy of 85.7% (95% CI 79.6–90.5, p < 0.0001). The algorithms by Arruda, Milstein, and Fitzpatrick yielded a predictive accuracy of 53.2%, 65.6%, and 44.7%, respectively. At comparable resolutions, the locAP AI achieved a predictive accuracy of 95.0%, 94.9%, and 95.6%, respectively (p < 0.001 for differences). (4) Conclusions: Our AI-based algorithm provided excellent accuracy in predicting the correct AP location. Remarkably, this accuracy is achieved at an even higher resolution of possible anatomical locations compared to established tree-based algorithms.