Customers rely more on online reviews than other sources while making decisions. However, the growth in fake or computer-generated reviews is threatening the validity of these reviews. The reason for this is that these reviews are hard to detect since they resemble the original ones. We utillize transformer based models Such as XLnet and BERT. These models were good at understanding the context or meaning in text and therefore led to improved identification of fine differences between actual versus computer-generated reviews. Results showed that BERT model gained 95.23% accuracy, while XLnet model achieved 94.80% accuracy in distinguishing between two types of reviews, and an ensemble model gave 96%, making it very reliable. Overall, our work makes online reviews more trustworthy, helping customers confidently reach conclusions on genuine feedback.