GPTZero is a classification model that predicts whether a document was written by a large language model, providing predictions on a sentence, paragraph, and document level. GPTZero was trained on a large, diverse corpus of human-written and AI-generated text, with a focus on English prose.