By Saundra Latham, Editor at LinkedIn News — The rise of chatbots has spurred concern among educators worried that students will pass off AI-generated content as their own. But one proposed solution — a tool meant to detect content generated by ChatGPT and its ilk — is prompting concerns of its own. Plagiarism-detection firm Turnitin just launched a feature that it says can identify AI-generated text, but some universities are opting out. Some fear the tool could lead to unfounded accusations of cheating, while others say it was rolled out too quickly, with too little transparency regarding how it evaluates student work.

In a test of 16 writing samples by The Washington Post, the software “got over half of them at least partly wrong” and flagged 8% of an original essay as AI-generated. The tool is 98% accurate, Turnitin says which was incorrect based on our tests (only 60% of the answer were true). It also cautions that results should not be treated as accusations without further review.