Ghostbuster is a model-agnostic detector for identifying text generated by artificial intelligence. It works by running documents through a series of smaller language models and performing structural search on possible feature combinations. Then, it trains classifiers on the selected features to determine whether the target document was generated by AI. Ghostbuster's training data includes news articles, student essays, and creative writing, but it may not represent all writing styles or topics. The data primarily consists of British and American English texts.