Comment by gs17
2 days ago
How well does that actually work in practice? I was using the demo on Fakespot's site and verbatim output from ChatGPT seems to get detected pretty well, but random HN comments got a lot of (presumably) false positives. "AI detectors" are kind of a crapshoot in my opinion, but I'm curious.
It's my belief that some people run their authentic HN comments through ChatGPT to "fix" their grammar. I've seen it especially with English language learners who are insecure about their abilities.