Grad-CAM is an Explainable AI algorithm that visualizes the regions of an input image most relevant for a convolutional neural network’s classification by leveraging the gradients of a target class flowing through the network’s last convolutional layers, helping to interpret model predictions. Using Grad-CAM can enhance explainability in AI classification applications, which is increasingly required under new European Union regulations. The aim of this study was to identify the factors influencing the visualization of the Grad-CAM algorithm to ensure that the visualization is as interpretable for experts in the context of forensic sexology. The study was conducted with the approval of the ethics committee. The study participants, prosecutors, were presented with an image and asked to assess the degree to which the image was explainable. Each participant evaluated 50 images, all of which contained exclusively pornographic content. The results of a one-way ANOVA, where the one variable corresponded to one of four color classes of the image and the second variable referred to the estimated degree of explainability with dynamic Grad-CAM modification, were statistically significant (F = 12.425, p < 0.001). The results show that dynamically modified Grad-CAM images were seen as more explainable in the classification process.