Large language models (LLMs) have dramatically reshaped the landscape of translation practice, offering unprecedented speed and scalability across hundreds of language pairs. Yet the enthusiasm surrounding these systems has frequently outpaced a sober accounting of what they cannot do. This paper examines the structural limitations and embedded biases of LLMs when they are applied to cultural translation—understood here not merely as the conversion of words between languages but as the negotiation of meaning across culturally situated contexts. Drawing on recent empirical and theoretical scholarship, the paper argues that LLMs reproduce and, in certain conditions, amplify the cultural asymmetries already present in their training data; that they systematically underperform on figurative language, politeness conventions, and minority or low-resource languages; and that the growing integration of AI into professional translation workflows raises questions about epistemic authority that the field has not yet adequately addressed. The paper concludes by suggesting that the most productive path forward is not the wholesale rejection of AI-assisted translation but a theoretically informed recalibration of its role—one that keeps human cultural judgment at the center of the process.

Chenxi Ban, Shu Wu. Cultural Translation in the Age of Artificial Intelligence: Limitations and Biases of Large Language Models. Academic Journal of Humanities & Social Sciences (2026), Vol. 9, Issue 3: 32-38. https://doi.org/10.25236/AJHSS.2026.090305.

[1] Jiménez Crespo, M. A. (2025). "If students translate like a robot…" or how research on human-centered AI and intelligence augmentation can help realign translation education. The Interpreter and Translator Trainer, 19(3–4), 277–295.

[2] Zheng, Y. (2025). On the underlying logic of translating Chinese texts on emergency management into English in the AI context. Scientific and Social Research, 7(12), 251–260.

[3] Tannous, B., & Haider, A. S. (2025). Assessing the accuracy of AI tools (Google Translate and Gemini) in translating Arabic idiomatic expressions and proverbs into English. International Journal on Artificial Intelligence Tools, 34(04–05).

[4] Alazzam, T. S., Alzghoul, M. A., & Alzghoul, R. M. (2025). Exploring AI's capability in translating English metaphors into Arabic. Theory and Practice in Language Studies, 15(7), 2319–2325.

[5] Matan, P., & Velvizhy, P. (2025). A neuro-symbolic AI approach for translating children's stories from English to Tamil with emotional paraphrasing. Scientific Reports, 15(1), 20348.

[6] Sofian, E., Nababan, M., Santosa, R., & Djatmika. (2025). Translating politeness: Mapping the characterisation of Klara the AI from Klara and the Sun and its Indonesian translation. Forum for Linguistic Studies, 7(11).

[7] Heathco, G. J. (2025). AI translated task setup for non-native English speaking students. Communication Teacher, 39(1), 26–32.

[8] Andalib, S., Spina, A., Picton, B., Solomon, S. S., Scolaro, J. A., & Nelson, A. M. (2025). Using AI to translate and simplify Spanish orthopedic medical text: Instrument validation study. JMIR AI, 4, e70222.

[9] Desrine, B. (2025). Translating the Indo-Caribbean in the AI era: The case of Trinidadian Bhojpuri. Archipélies, (19), 114–146.

[10] Peterson, D. J. (2025). Translating Franz Kafka's "Josefine, the Singer or the Mouse People" as a metaphor for AI. Humanities, 14(2), 21.

[11] Buckley, A. O., O'Sullivan, B., McKean, C., & Frizelle, P. (2026). An AI tutorial for speech and language therapists: Translating concepts from the AI literature into accessible knowledge and clinically relevant applications. International Journal of Language & Communication Disorders, 61(2), e70201.

[12] Tao, H. (2025). Google Translate: AI in interpreting and translating for speech-language pathology. Journal of Clinical Practice in Speech-Language Pathology, 27(2), 215–216.