Language Nuances: usually, AI detectors stumble when working with nuanced human language use, like irony or humor, resulting in Fake positives or negatives. Perplexity is a measure of how predictable a bit of text is. https://www.brodneil.com/undetectable-ai/