Delamination is a critical failure mechanism of laminated composites, consisting of the separation of two adjacent laminas. This can occur both in the case of static and fatigue loadings. Designing against this failure mode requires the use of advanced numerical tools, the two most common of which are the Virtual Crack Closure Technique (VCCT) and the Cohesive Zone Model (CZM) [1]. VCCT computes the local value of the strain energy release rate G at the delamination front. It is based on linear elastic fracture mechanics, thus preventing any process zone modelling. The technique is known to be mesh dependent because of a discontinuous node-based description of the delamination front. CZM is based on the use of cohesive elements that implement a non-linear traction separation law. These elements can model effectively the process zone ahead of the delamination front, which is described continuously via the elements. Comparison of both approaches can be found in the literature [1-3]; however, all comparisons were performed at the global level, by comparing the load-displacement curve of the simulated structure or specimen. A local comparison of the computed values of G and its decomposition in the different modes was so far not possible because. This was especially true for large and curved fronts, for which the VCCT shows unreliable readings. Therefore, a clear understanding of the real different and analogies of the two techniques at the local level is still lacking. In this work, two innovative methods are used together to perform a local comparison of the two approaches. First, a smoothing algorithm is used to allow a more coherent description of the VCCT delamination fronts, thus allowing more reliable and continuous readings of G [4]. Moreover, a recent cohesive elements formulation implementing the J-integral is used to accurately compute the local G in different locations of the process zone [5]. The comparison is thus performed on a non-standard EndNotched Flexure (ENF) specimen, presented in [5] and shown in Fig. 1a. This specimen is characterised by a large and curved delamination front upon loading. To investigate the effect of the process zone in this comparison, two process zone lengths were investigated. The comparison is performed by extracting isodamage curves in the process zone of the CZM simulation (the dissipated energy wd is considered as damage [5]); these curves are used as delamination fronts for VCCT simulations. In these simulations, the delamination propagation was prevented by imposing an infinite fracture toughness, so that any value of G could be read. For the exemplificative case of an applied displacement of 28 mm, Figure 1b shows the extracted delamination fronts; Figure 1c shows a comparison between the VCCT and the CZM without the smoothing algorithm for the VCCT simulations, highlighting how this comparison was so far impossible; Figures 1d-f show the comparison for three iso-damage lines. As shown, the predicted trends are similar, but the absolute values often show differences. In addition, the significative differences appear to affect mainly GIII, while GI and GII are more similar: this implies that the total GTOT (the sum of the three components) is also different for the two techniques. A different GTOT necessarily implies a different delamination propagation prediction. These differences are attributed mainly to the presence of the fracture process zone, which thus is critical in controlling the propagation of this delamination case.