Pangram AI Detector records false positive rate below 1 in 10,000 on thousands of pre-2020 human passages and 3,400-word fiction samples

@MushtaqBilalPhD No, most people are reading the essay and seeing an essay densely packed with common AI phrases and structures. Pangram doesn't catch all AI outputs, but it has a false positive rate of < 1 in 10,000. We don't say that something is AI unless we are sure of it (esp. >100 words)