Researchers find six patterns that make AI text so exhausting to read

Researchers at New York University and the University of Texas at Austin identified six recurring linguistic patterns in language model outputs that make AI-generated text difficult for humans to process. The study suggests these issues stem from training models on code and mathematical data, which prioritize logical correctness over readable prose. This finding explains the common fatigue users experience when consuming AI writing and highlights a fundamental misalignment between current training objectives and human-centric communication standards.