Amazon is a major e-commerce platform and one of the primary online destinations for purchasing electronics. On such platforms, sellers use product titles to provide key descriptions of their items, making these titles the first elements that customers read and rely upon when deciding to buy. However, the growing use of automated text-generation tools has made it easier to create Arabic product titles that resemble genuine listings but may not accurately describe the product. These tools are often used to save time and reduce manual effort, as a large variety of products makes manual title writing time-consuming. This practice can reduce customer trust when the generated titles fail to reflect actual product properties. This study develops a machine learning model for detecting Arabic synthetic product titles on Amazon. The objective is to construct a dataset of real and synthetically generated Arabic smartphone titles and train a classifier capable of distinguishing organically written titles from template-generated synthetic texts. Real titles were collected from Amazon.sa, and synthetic titles were produced using a structure-aware template approach. A BERT-based binary classifier was fine-tuned, achieving training and testing accuracies of 99.1% and 98.3%, respectively, demonstrating strong in-domain performance. Synthetic titles were generated using a structured template-based pipeline rather than neural generative language models. Therefore, the findings are limited to template-generated synthetic content. This study provides a controlled framework for evaluating synthetic title detection in structured Arabic e-commerce contexts.