Hello Paula!

Thanks for reaching out and for your close attention to how Turnitin’s AI writing detection is performing in your classroom. 

• Turnitin is committed to maximizing the effectiveness of our AI detection system while minimizing the risk of false positives. 
• Our goal is to keep the false positive rate under 1% for documents with over 20% AI-generated content. In practical terms, that means only about 1 in 100 fully human-written documents might be mistakenly identified as AI-written.
• To validate this, we conducted further testing in April 2023 on 800,000 academic papers written before the release of ChatGPT. These tests confirmed our false positive rate remains below 1%.
• To maintain this high level of precision, there is a trade-off: we may miss up to 15% of AI-written content in a given document. For example, if we flag 50% of a paper as AI-generated, the actual amount could be closer to 65%. We accept this margin because we prioritize accuracy and fairness to students, ensuring that human writing isn’t wrongly flagged.

We continuously adapt our model based on real-world submissions and the evolving landscape of large language models. You can read more in our AI detection FAQ here: Turnitin AI Writing Detection Capabilities

If you have specific examples you'd like our team to review, we recommend submitting them via the “Feedback” button located in the AI writing report itself. This directly supports ongoing improvements to the model.
 

Please feel free to follow up if you have more questions. We’re always here to continue the conversation  

- Kat, Turnitin Team