ChatGPT creator OpenAI announced a new tool that can differentiate between text written by AI and that written by a human. The tool, called AI Text Classifier, could help teachers detect student essays written using ChatGPT. It is free for users with an OpenAI account. According to OpenAI, the classifier has a 26% success rate when it comes to identifying AI content, making it not especially accurate . To utilize the tool, users paste in text, and the AI program will label it as "very unlikely," "unlikely," "possibly," or "likely" that the words were created using AI. It also has a fifth category saying it's "unclear if the text is" AI-generated. The tool, described as "a fine-tuned GPT model," requires at least 1,000 characters, or 150 to 250 words. OpenAI acknowledged the tool's limitations , saying it's best used in combination with other methods to determine if text is written by an AI. The tool's release comes as more school districts are banning students from using ChatGPT over plagiarism and accuracy concerns. Also this week, Stanford researchers announced DetectGPT to identify content created by large language models like ChatGPT. Other AI writing detectors include GTLR from MIT's Watson AI lab, Packback's tool, Turnitin software, and AI Detector Pro .