Cardiac computational models often rely on late gadolinium magnetic resonance imaging (LGE MRI) to identify cardiac anatomy and substrate information such as scar and fibrosis. Identification of tissue fibrosis from LGE MRI is an especially difficult task due to the lack of robust ground truth data. Furthermore, most contemporary modeling approaches discretize the tissue into either fibrosis, healthy, or scar with sharp cutoff values to which the subsequent simulations are highly sensitive. These modeling approaches simplify the physiological transition from healthy to fibrotic tissue with a range of intermediate fibrosis states in between. In the present study we aimed to explore the influence of a fibrosis gradient on arrhythmogenic potential in patient specific computational models. We mapped LGE MRI intensity to the degree of fibrosis within a patient specific computational model in a number of discrete steps and compared in-silico arrhythmia induction in these models and a traditional model with only a single fibrotic cutoff. We find that using an increasing number of steps to represent the transition between healthy and fibrotic tissue results in more prevalent, more complex, and somewhat more stable presentation of induced arrhythmias. This includes the ability of the model to sustain multiple simultaneous reentrant wavefronts when the gradient of fibrosis is implemented, a feature that is not common among contemporary modeling approaches and better replicates the complex multi focus activity of atrial fibrillation. This representation of fibrosis represents a more comprehensive leveraging of the rich data available in LGE MRI while keeping the complexity and computational cost of the models comparable to traditional approaches and results in more complex arrhythmia simulations.