Abstract Background Ablation of persistent atrial fibrillation (AF) beyond pulmonary vein isolation (PVI) remains challenging, despite increasing numbers of eligible patients. Previous studies suggest a possible role for non-PVI lesions targeting atrial drivers, but their identification can be difficult and subject to inter-operator variability. Purpose We investigated feasibility for an electrogram-based artificial intelligence software system to assist in the identification of left or right atrial dispersion sites as ablation targets for persistent AF using the penta-spline multi-electrode pulsed-field ablation (PFA) catheter, a recent technique which is more conventionally employed for PVI. Methods This is a single-centre retrospective descriptive study of a cohort of patients with persistent AF, who underwent their first-time ablation from May to November 2023 using the Farawave PFA catheter (Boston Scientific) and the VX1 system (Volta Medical). This system relies on off-line pre-training of multiparametric machine learning algorithms on a database of annotated AF intracardiac electrograms. It connects to cardiac electrophysiology recording systems via a custom data cable to receive analogue data and digitize them through an integrated converter. Its algorithms analyse signals in the cardiac chamber of interest to adjudicate the local presence or absence of dispersion and provides relevant visual cues to operators in order to target energy delivery. Results We identified 16 patients who underwent PFA for persistent AF (long-standing in 5/16, 31.3%), including PVI and further ablation guided by VX1. See table 1 for baseline population characteristics and table 2 for procedural characteristics. All underwent ablation with uninterrupted anticoagulation, general anaesthesia, use of the Rhythmia electroanatomic mapping system and its Orion mapping catheter (Boston Scientific) to feed signals into the VX1 system. VX1-guided electrogram localization and ablation mostly targeted the anterior region of the left atrium (16 segments), followed by its septal region (6 segments) and its lateral region (4 segments); the median number of applications per procedure was 88 [IQR 68-108], including a median of 76 [61-85] in the left atrium and 12 [0-15] applications in the right atrium. The median procedure duration was 123 [114-137] minutes; the median fluoroscopy time was 32 [26-40] minutes and the median radiation dose was 12.5 Gy∙cm2 [10.0-25.8]. Sinus rhythm was restored either during ablation or by electrical cardioversion (see Table 2). Two complications occurred one femoral arteriovenous fistula requiring stenting and one tamponade requiring drainage. At a three-month follow-up, 87.5% did not present AF recurrences. Conclusion Real-time detection and annotation of electrograms to track dispersion using AI-assisted technology to support PFA with a multipolar catheter is feasible. The interest of the approach should be validated by further studies.Table 1.Table 2.