Your AI-detector needs a calibrate, I think. What made you think it was llm-written?