Terminology translation in the generative AI era

Abstract The rise of generative artificial intelligence (GenAI), especially large language models (LLMs), is reshaping both the practice and theory of terminology translation. Traditionally grounded in equivalence, standardization, and concept-term correspondence, terminology translation now confronts systems that generate context-conditioned outputs shaped by pre-training, instruction tuning, and multimodal capabilities. This paper examines how GenAI supports multilingual term generation, cross-lingual semantic alignment, terminological variation management, and multimodal knowledge representation, while also exposing persistent risks of hallucination, data bias, domain-specific inaccuracy, and reduced translator agency. Theoretically, the paper argues that GenAI destabilizes static models of term equivalence and calls for a more dynamic account of concept-term relations in computationally mediated environments. It therefore proposes a Dynamic Mediation Framework (DMF), which understands terminology translation as an iterative, context-sensitive process of knowledge negotiation in which human expertise and machine capability operate in complementary roles. By linking practical innovation with epistemological reflection, the study contributes to the reorientation of terminology translation research in the GenAI era.