Can ChatGPT trick Turnitin software? Can Turnitin detect AI generated work?

In a recent experiment, the viability of Turnitin, a widely used software in educational institutions to detect plagiarism, was tested to see if it could identify AI-generated content. The experiment involved submitting essays on various scholarly topics, composed by ChatGPT, to Turnitin to determine if the software could distinguish between human-written and AI-generated work.

Six essays were submitted, covering topics such as cybernetics, qualitative and quantitative research, Carl Jung's psychology, Teach back in cybernetics, and solar panel manufacturing. The results varied in how accurately Turnitin was able to identify AI-generated content. While Turnitin successfully flagged essays that were entirely AI-generated, there were instances where it missed certain sections. In a final test involving interleaving of AI-generated content with manually written text, Turnitin was able to detect some AI-generated portions but not others.

This experiment showcases the capabilities and limitations of Turnitin in detecting AI-generated work, raising questions about the effectiveness of such software in the face of advancing AI technology.

Keywords

Turnitin, ChatGPT, Plagiarism detection, AI-generated content, Educational technology, Scholarly essays

FAQ

Can Turnitin accurately identify essays that are entirely generated by ChatGPT?
How did Turnitin perform in detecting AI-generated content in scholarly essays on specific topics?
What are the limitations of Turnitin in discerning between AI-generated and human-written work?
How does interleaving human-written text with AI-generated content impact Turnitin's detection capabilities?
What implications does this experiment have for plagiarism detection software in educational settings?