Top AI models write sentences that are harder for people to process.
Linguists at NYU and the University of Texas posted a study on Wednesday comparing nine leading models with human writing. All nine used rarer word combinations than people do, a pattern known to slow readers down. The authors suspect training that rewards short, dense answers, and they found Claude Opus 5.5 easier than Opus 5.
Why it mattersIf AI answers feel tiring to read, the cause may be the model, not you. Asking a chatbot for plain, short sentences is a cheap way to push back.
Read in the app — free, in 9 languages
Related stories
Mathematicians set out to rewrite OpenAI's 719 AI-made papers.2026-10-25Claude designed proteins that latch onto deadly scorpion toxins.
2026-10-24OpenAI's AI found a flaw in the code that shrinks web traffic.
2026-10-24ChatGPT drew fake scandal photos of US candidates 80% of the time.
2026-10-24Claude proved that 123 code libraries are free of memory crashes.
2026-10-24