<p>Before making a purchase, consumers often look for reviews of the product or service online. These reviews also help businesses since they provide valuable insight into how customers perceive their goods and services. Having trustworthy information was especially important during the COVID-19 pandemic, when people were cooped up at home and had to rely on internet reviews to pass the time. Not only did the quantity of reviews increase during the pandemic, but so did the context and preferences. Reviewers of spam take note of these shifts and work to refine their deceitful methods. For financial gain or competitive advantage, spam reviews often include false, deceptive, or fraudulent statements meant to mislead consumers. Further, dealing with different languages makes fake review identification more complex, especially in the absence of labeled data for model training. Therefore, our study suggests a transformer ensemble model that uses a large language model (LLM) to translate foreign languages into English. We turned to two neural machine translation models-Google Translate and LibreTranslate-to translate two languages: Arabic, Spanish. Next, a set of pre-trained models was used to examine the tone of the reviews. These models included GPT-3, an LLM from OpenAI, bert-base-multilingual-uncased, and Twitter-Roberta-Base. Based on our experimental results, it is evident that a foreign language fake review is achievable through English translation. The suggested ensemble model outperforms both the independent pre-trained models and LLM, and the accuracy and f1-score of fake review detection on translated reviews are 90% and 87.6%, respectively, on the Arabic language. To the best of our knowledge, our proposed framework is the first that use an LLM model for Arabic and Spanish languages.</p>