This article investigates the comparative linguistic features of artificial intelligence (AI)-based translation systems, focusing on how machine translation models render linguistic structures across typologically different languages, particularly Uzbek and English. The study analyzes syntactic, lexical, semantic, and pragmatic shifts occurring in AI-generated translations and compares them with human translation norms. Special attention is given to neural machine translation (NMT) systems and their ability to handle idiomatic expressions, polysemy, and word order variations. The research highlights both the strengths and limitations of AI translation tools, emphasizing issues such as loss of cultural nuance, structural simplification, and contextual misinterpretation. The findings suggest that although AI translation systems have significantly improved in fluency and accuracy, they still require linguistic refinement to fully capture deep structural and cultural meanings across languages.
Esanova et al. (2026) studied this question.