Abstract
<jats:p>The increasing use of artificial intelligence (AI) in academic translation has raised important questions regarding the quality of meaning transfer beyond surface-level fluency. While AI-generated translations often demonstrate grammatical accuracy and lexical naturalness, less attention has been paid to how contextual meaning and implicature are reconstructed in comparison with human translation. This study investigates contextual framing and implicature transfer in human and AI-generated translations of academic abstracts. Drawing on pragmatic theory and functionalist approaches to translation, the research adopts a qualitative comparative design grounded in linguopragmatic analysis. A corpus of 10–15 academic abstracts was translated independently by human translators and an AI-based system, and instances of implicature were identified and analysed in relation to contextual adequacy. The findings reveal systematic differences between translation modes. AI-generated translations frequently exhibit contextual framing shifts, implicature loss, and implicature strengthening through explicitation. Although linguistically fluent, such translations tend to approximate contextual meaning through probabilistic association rather than inferential reconstruction. Human translations, by contrast, demonstrate greater sensitivity to disciplinary positioning, epistemic calibration, and discourse coherence. The study argues that translational adequacy in academic contexts is fundamentally grounded in pragmatic reconstruction rather than formal equivalence. These findings contribute to translation theory and provide critical insight into the limitations and appropriate use of AI-assisted translation in scholarly communication.</jats:p>