Atrial fibrillation (AF) is the most common cardiac arrhythmia globally and can cause stroke, heart failure, and dementia. Catheter ablation for AF has emerged as an effective rhythm-control strategy for AF. However, even the gold standard, pulmonary vein isolation (PVI), has s single procedure success for persistent AF of ~60%. This warrant the optimisation of ablation strategy. Fibrosis-substrate isolation ablation (FISA) is a promising new ablation strategy currently showing success in clinical trials. However, the left atrial (LA) fibrosis border zone (FBZ) tissue needs to be characterised to perform FISA. This study investigates the impact of FBZ characterisation on in-silico FISA outcomes for AF models. The data used in this study were LA surface meshes obtained from late-gadolinium enhanced magnetic resonance images (LGE MRI) from 10 patients. These surface meshes were then thresholded using MRI image intensity ratio into healthy tissue, dense fibrotic tissue and four levels of FBZ. Patient-specific LA meshes were combined with Courtemanche atrial electrophysiology models to created 10 LA models. Each model was then simulated with two AF scenarios for 10s, where one AF scenario was reentry initiation around the left superior pulmonary vein and other AF was reentry initiation on the anterior of the LA. After 10s, FISA was performed at each FBZ level and AF termination was evaluated 2s post-ablation. Our results showed that the success of AF termination was increased when FISA isolated a large amount of FBZ tissue within its ablation lesions (7/10 patients when ablating FBZ levels 1 and 2). In addition, we also showed that FISA is more effective for patients with Utah fibrosis stages III & IV (>20% fibrosis) and less effective for stages I & II. These results can help clinicians to improve stratification of AF patients and implementation of FISA strategy.