The cohesive zone model (CZM) has found wide acceptance as a tool for the simulation of delamination in composites and debonding in bonded joints and various implementations of the cohesive zone model dedicated to fatigue problems have been proposed in the past decade. In previous works, the authors have developed a model based on cohesive zone to simulate the propagation of fatigue defects where damage acts on cohesive stiffness, with an initial (undamaged) stiffness representative of that of the entire thickness of an adhesive layer. In the case of a stiffness that is order of magnitude higher than the previous one (for instance, in the simulation of the ply-to-ply interface in composites), the model prediction becomes inaccurate. In this work, a new formulation of the model that overcomes this limitation is developed. Finite element simulations have been conducted on a mode I, constant bending (constant G)-loaded double cantilever beam (DCB) joint to assess the response of the new model with respect to the original one for varying initial stiffness K₀ and cohesive strength σ₀. The results showed that the modified model is robust with respect to changes of two orders of magnitude in initial stiffness and of a factor of two in σ₀.