Peer Review Report For: Translating dialects between ChatGPT and DeepSeek: Yemeni San’ani Arabic terms as a case-in-point [version 2; peer review: 1 approved, 1 approved with reservations]

Background This study aims to detect the efficiency of two Artificial Intelligence (AI) translation models, ChatGPT and DeepSeek, in the translation of Yemeni San’ani Arabic (YSA) dialectical terms into English. As dialectal Arabic presents significant linguistic variability and cultural specificity, accurate translation remains a major challenge for the current ChatGPT and DeepSeek (and perhaps other AI models). Methods Fifty San’ani Arabic terms were involved in the translation process, assessing the ability of both models to capture their semantic fidelity, cultural relevance, and contextual accuracy. Results The study findings reveal that, while both models demonstrate a foundational understanding of Standard Arabic (SA), their performance diminishes considerably when faced with the nuances and idiomatic expressions of the San’ani Arabic dialect. ChatGPT displays a relatively better performance in certain cases, particularly when translating terms with dialectical connotations. However, both models exhibit limitations, such as literal translation, misinterpretation, or complete ignorance of the intended meaning. Conclusions The study concludes by highlighting the critical need for dialect-aware AI development and provides recommendations for improving the dialectical accuracy and cultural sensitivity of AI model translation.