Idioms, being culturally bound linguistic units, pose significant difficulties in comprehending them, especially when linguistic structures differ between the source and target languages. This study investigates the precision of artificial intelligence (AI) tools in decoding idioms. It also examines whether the efficiency of AI translation tools (AITTs) differ in decoding idioms across the target language it translates into (Arabic into English or English into Arabic). Through a comprehensive analysis of 10 idiomatic expressions within their contexts to the translation by five AI tools, i.e., ChatGPT, QuillBot, Copy AI, Poe, and Gemini. The results of the study showed that there are potential problems in the process of decoding idioms from English into Arabic and vice versa. The finding showed that AITTs' level of proficiency differs. ChatGPT ranked first while QuillBot ranked last. Yet, their level of proficiency is still weak. The findings also showed that AITTs' level of proficiency is better when rendering from Arabic into English to translating form English into Arabic. This study contributes to the AITTs by providing insights into the capabilities of these tools for translation accuracy.