AI & Enterprise
AI-favoured words spread to humans, confusing detectors
A study said AI-style wording is seeping into human-written and edited text, potentially undermining methods that judge AI authorship by word choice alone. A computational linguist analysed 18,989 abstracts on arXiv from 2019 to 2026 and compared the use of terms often criticised as AI-generated, such as “delve” and “intricate”. Usage rose sharply in computational linguistics and neuroscience after late 2022, while probability theory changed little.
