A recent study published on arXiv investigated whether professional translators could identify AI-generated text. In an experiment involving 69 translators assessing short stories, a significant minority (16.2%) were able to distinguish AI-written stories from human-authored ones. However, a similar number misclassified the texts, sometimes due to a preference for AI-generated content. The research identified low burstiness and narrative contradictions as key indicators of AI authorship, while grammatical accuracy and emotional tone were less reliable. AI
Summary written by gemini-2.5-flash-lite from 1 source. How we write summaries →
IMPACT Suggests potential challenges for editors and readers in distinguishing human from AI-generated content.
RANK_REASON Academic paper on AI text detection capabilities. [lever_c_demoted from research: ic=1 ai=1.0]