AI Writing Detectors Fail to Distinguish Between Human and AI Authors
OpenAI confirms that AI writing detectors, including their own discontinued AI Classifier, do not reliably distinguish between AI-generated and human-generated content. These detectors often yield false positives and can be defeated by rephrasing. OpenAI also clarifies that ChatGPT itself cannot determine if text is AI-written or not and warns about the model's propensity to confabulate false information. While humans can sometimes detect AI writing through changes in style or tell-tale signs, it is safest to avoid automated AI detection tools as they have high false positive rates.