Authors Guild test finds some AI detectors perfectly identify human writing while others fail on every single text

The Authors Guild recently evaluated five AI detection tools, finding significant inconsistencies in their ability to distinguish human-written content from machine-generated text. While some programs correctly identified all human writing samples, others incorrectly labeled them as AI-generated in every instance. These findings highlight the unreliability of current detection software, particularly as professional writing styles increasingly mirror the statistical patterns used by large language models. This ongoing uncertainty complicates efforts to verify authorship in academic and professional settings.