Every year many people die in road accidents, however, advance in Intelligent Transportation Systems (ITS) is possible to avoid it. The environmental conditions make it difficult to perform accurate and reliable traffic sign recognition (TSR), especially the recognition of Arabic Traffic Signs (ATS). Conventional recognition methods based on characteristics face constraints, while deep learning (DL) models have shown better performance. In this study, we introduce an Arabic TSR dataset tailored for Explainable Artificial Intelligence (XAI) applications, emphasizing transparency and interpretability. We evaluated the performance of YOLOv8 and SSD (Single Shot Multibox Detector) on this dataset. YOLOv8 achieves mAP@0.5 of 94.20%, significantly outperforming SSD at 88.37%, and also excels in recognition speed. In addition, we implement Grad-CAM analysis for YOLOv8 to visually illustrate the model image highlighting process as it learns which features in ATS are essential under different operating conditions. The results of this study show that YOLOv8 outperformed other transport solutions in terms of accuracy and interpretability, showing that XAI techniques have the potential to substantially improve the reliability and transparency of the solutions deployed in ITS.