Turnitin AI detection false positive risks

Turnitin has integrated an artificial intelligence (AI) writing detection tool which helps identify written submissions that may have been generated by AI. Read more about how to be mindful of the risk of false positives.

What is it?

Turnitin has a new AI writing detection tool that helps flag text that may have been generated by AI. It appears as a percentage in an AI indication button and is not visible to students. Since the implementation of the tool, Turnitin has emphasised that the percentage shown in the AI writing detection indicator and AI writing report should be used with caution. The tool does not of itself indicate academic misconduct and further review of the submission should be undertaken.

Updated guidance in interpreting and utilising the tool

The Turnitin tool provides an overall percentage that represents the proportion of the work that the tool is confident was written by generative AI (with 98 per cent confidence). It also highlights the text that it believes was AI generated. This is separate from the text matching for the Similarity report. More information about using the tool is available in LEO guides under ‘Detection indicator’.

Turnitin recently provided the following update:

“Our testing has found that there is a higher incidence of false positives when the percentage is between 1 and 20. In order to reduce the likelihood of misinterpretation, the AI indicator will display an asterisk (*) for percentages between 1 and 20 to call attention to the fact that the score is less reliable."

This tool may be useful as an initial flag to indicate that generative AI may have been used. However, due to the risks of false positives, it should not be used as the sole evidence to forward an allegation of potential academic misconduct. Also note that there are false negatives, so a low percentage does not necessarily mean that there is no AI generated text.

What action should I take?

Under ACU’s Student Academic Integrity and Misconduct Policy, unauthorised use of generative AI in work submitted for assessment is a form of academic misconduct. Due to the risk of false positives and negatives, this tool should be used in conjunction with other features of the submitted work that may indicate unauthorised use of AI, including:

• fabricated references
• fabricated quotes
• incorrect information
• the material being slightly off topic, repetitive or has the wrong focus.

For a more comprehensive list of suspicious features, please view the Academic Integrity FAQs on the CEI webpage.

If you see work that Turnitin flags as being AI generated, always look for other features that may indicate AI has been used. We advise that extreme caution is used when the percentage indicator is less than 20 per cent, and that cases are only referred where there is very strong independent evidence that the work is not the student’s own.

Further resources

• LEO guides have been updated to provide guidance.
• Turnitin have provided an AI writing detection resource and FAQs.
• A recent Staff Bulletin article provided suggestions for assessment design to help mitigate the impact of generative AI tools and summarised assessment design and feedback resources available to staff at ACU.
• Workshops regarding the best use of Turnitin to detect possible academic misconduct will be available in the coming weeks.