This study aims to explore differences in the translation outputs of three artificial intelligence (AI) platforms, namely ChatGPT, Gemini, and Claude AI, in translating Mahmoud Darwish’s poem “Ilā Ummī” from semantic and cultural-aesthetic perspectives. The study employs a qualitative method with a descriptive-comparative design. The primary data consist of the original text of the poem “Ilā Ummī” from Al-A'mal al-Ula, Volume 1, while the translation data were obtained by providing the three AI platforms with the same translation prompt. Data analysis was conducted through three stages: data reduction, data presentation, and conclusion drawing, guided by the principle of aesthetic-semantic balance. The findings indicate that the three AI platforms exhibit distinct characteristics and stylistic tendencies. ChatGPT tends toward aesthetic expansion, producing Indonesian sentence structures that are fluent, poetic, and rich in expressive value, although it occasionally reduces complex metaphors. Gemini occupies a more moderate position, tending to simplify meanings to make them more accessible to target readers, although this approach occasionally results in semantic shifts. Claude AI, meanwhile, demonstrates a high degree of semantic accuracy and a tendency to preserve the distinctive structures and theological diction of the source text, although its translations tend to be more rigid in simple, domestic lines. These findings suggest that the three platforms not only produce translations with different degrees of semantic equivalence but also display distinctive aesthetic tendencies in processing literary texts. In the context of computational literary translation, the principle of aesthetic-semantic balance is an important consideration in ensuring that the transfer of meaning into the target language retains the poetic character of the source text.
Copyrights © 2026