Turnitin is a well-known tool for educators that scans student papers and matches passages to its extensive database of existing content to flag instances of plagiarism. It recently added a feature that allows it to detect whether students use AI writing tools such as ChatGPT in their assignments.

It’s still early days, but feedback from turnitin ai detection over the first six weeks of using Turnitin ai detection has led to a few tweaks. Chief Product Officer Annie Chechitelli explains what’s been learned so far in a new blog post.

Educators need to know that even though Turnitin’s new AI detection software has a less than 1 percent rate of the most problematic kind of error (false positives, where real student writing gets incorrectly flagged as cheating), it’s not foolproof. And that’s important, because if students are falsely accused of plagiarism, they could lose out on academic opportunities.

AI-Powered Academic Integrity: Exploring the Role of Turnitin’s AI Detection in Preventing Plagiarism

The AI writing detection model works best with long-form prose, such as essays and research reports. It can’t reliably detect non-prose text, such as poetry, scripts, code, ordered/unordered lists, or tables. Chechitelli also notes that the accuracy rate of detecting AI writing at the document or sentence level can vary, depending on how similar it is to existing sources in Turnitin’s database.

Finally, the algorithm is designed to recognize a wide range of language models—including GPT-4—and will continue to evolve as new versions emerge. This will ensure that the algorithm stays ahead of advanced generative AI models that may not yet be detected.