I was curious about how reliable these tools actually are, so I ran an experiment. I took an old essay I wrote before ChatGPT even existed—like, 2019 pre-AI era—and ran it through three popular AI writing detector websites. One said it was 2% AI-generated (basically human). Another said it was 27% AI-generated. A third flagged it as 68% AI-generated and said it was 'likely written by AI.' Same essay. Written by me, years before AI was a thing. Three completely different results. This is terrifying because professors are using these tools to make life-changing decisions about students' academic careers. If detectors can't agree on a human-written essay from 2019, how can anyone trust them? has comparisons showing different detectors give wildly varying scores. Has anyone else tested detectors on old work? What did you find?