Atrial fibrillation (AF) is the most common arrhythmia and is often treated with catheter ablation. However, many patients experience recurrence, and it remains unclear whether recurrence results from the incomplete elimination of the native arrhythmogenic substrate or new arrhythmogenic substrates created by ablation. We aim to investigate this using bi-atrial digital-twins (DTs) constructed from pre- and post-ablation MRI scans of 22 AF patients, half of whom experienced recurrence within one year. In the DTs, the atrial tissue was characterized as non-fibrotic, fibrotic, and, exclusively in the post-ablation left atrium (where lesions were delivered), as ablation-scar. Substrate arrhythmogenesis was assessed in DTs using a sequential pacing protocol from 40 pacing sites to induce reentries. For each DT, arrhythmogenic propensity was quantified by the potential reentry-sites (PRs) and the proportion of pacing-sites that induced a reentry (vulnerability). Only DT-pairs with at least one pre-ablation PR were included in the analysis. In DTs of patients whose AF did not recur (N-DTs; n=9), both the number of PRs and vulnerability decreased from pre- to post-ablation (ΔPRs=-2.7±2.2; Δvulnerability=-8.3±8.4%). In contrast, in DTs of patients whose AF recurred (R-DTs; n=10), the change in number of PRs was negligible (ΔPRs=-0.2±1.8; p=0.017) and the vulnerability increased post-ablation (Δvulnerability=9.0±17.2%; p=0.014). In both N-DTs and R-DTs, less than 10% of pre-ablation PRs persisted post-ablation. Most R-DTs exhibited 2-3 newly emergent PRs post-ablation, while most N-DTs exhibited only 1. In post-ablation DTs, reentries in the left atrium anchored around ablation scar. Scar-anchored reentries were observed in 7 R-DTs vs only 3 N-DTs and they were inducible from more pacing sites (5.3±4.7) than reentries not anchored around scar (2.0±1.7; p=2e-5), contributing to the heightened vulnerability of R-DTs post-ablation. These findings show that while ablation disrupts the native arrhythmogenic substrate, the lesions can create new arrhythmogenic substrates that could contribute to AF recurrence.