Frequent ChatGPT users are accurate detectors of AI-generated text (2025) (arxiv.org)

Full title (shortened due to submission limitations):

"People who frequently use ChatGPT for writing tasks are accurate and robust detectors of AI-generated text"

Basically, if someone is exposed to LLM writing they become very skilled at detecting it elsewhere. Fits my own experience imo. I’ve certainly have had false negatives where I’ve been fooled by the AI, but I’ve never once made a false positive.