This article examines the linguocultural losses occurring in literary translations generated by artificial intelligence (AI), using Abdulla Qodiriy’s novel *O‘tkan kunlar* (Days Gone By) and its German translations as case studies. The study aims to compare the translation of Uzbek linguocultural units into German via Google Translate and ChatGPT against a professional translation and to determine the extent to which national-cultural content is preserved. Selected *realia* (culture-specific terms), ethnographic units, kinship terms, and the micro-texts containing them served as the research material. The analysis employed criteria such as semantic equivalence, linguocultural adequacy, pragmatic alignment, and the preservation of national-cultural components, alongside techniques including generalization, concretization, transliteration, and explication. The results indicate that in Google Translate versions, the incorrect translation or generalization of certain units leads to a loss of cultural information, whereas ChatGPT demonstrates a tendency to preserve the national-cultural characteristics of *realia* by combining transliteration and explication. Micro-text analysis revealed that AI translations are relatively close to the original in terms of lexico-semantic and compositional structure, though they lack sufficient adaptation to the natural literary norms of the German language regarding certain syntactic constructions. Research results confirm the necessity of a comprehensive approach—incorporating linguocultural and pragmatic criteria alongside lexical-semantic correspondence—when evaluating the quality of literary translation using artificial intelligence.