Despite its shortcomings, the programme remains an essential resource for instructors attempting to detect AI-generated writing. The programme works by ranking the text’s likelihood of being created by an AI on a five-point scale ranging from “impossible” to “probable.” According to Axios, the programme is best successful on text samples of over 1000 words in English, while its efficacy degrades in other languages. The technology also cannot tell the difference between computer code authored by humans and AI.
