Can professional translators identify machine-generated text?

This study explores the ability of professional translators to identify AI-generated texts, specifically short stories produced by ChatGPT-4o. Out of 69 participants, only 16.2% accurately distinguished AI-generated content from human-written texts, with low burstiness and narrative contradiction being the most reliable indicators of synthetic authorship. The results highlight challenges in recognizing machine-generated text, suggesting implications for the editing and evaluation of AI-produced content in professional translation contexts.