Also, it'd be interesting how many pre-2020 papers their "AI detector" marks as AI-generated. I distrust LLMs somewhat, but I distrust AI detectors even more.