"Over four months, LLM users consistently underperformed at neural, linguistic, and behavioral levels."

Not surprising but damn.