A new research paper explores the structural and cognitive reasons why humans often find AI-generated language difficult to read and comprehend. By investigating the specific linguistic patterns that make machine-written text feel unnatural, the study aims to help developers improve the readability of future large language models.