Southampton and Newcastle universities have both said they aren’t renewing Turnitin, a similarity detection platform, going forward.
- University of Southampton’s announcement to move away from Turnitin
- Newcastle University statement on Turnitin
Although Newcastle cites a number of reasons, both institutions mention that the non-renewal is at least partly due to Turnitin’s intention to change its End User License Agreement (EULA) and add new terms allowing them to use submissions made by students to train their AI services. Both are planning to stop using it from next academic year. I suspect other UK universities will follow suit.
These decisions may be triggered by a specific supplier’s terms but they create an opportunity to ask a much bigger question: do we still need similarity detection?
The Guardian reported in June 2025 that it contacted 155 universities across the UK (131 responded) to request information about plagiarism and AI-related misconduct numbers – plagiarism numbers have been going down year-on-year since academic year 2021-22. AI related misconduct cases were increasing but not to the same extent as plagiarism cases have reduced.
A survey by Higher Education Policy Institute (HEPI) this year showed that of the 1054 UG students surveyed, 94% of the students have used AI to help with their assessments, jumping up from 53% in the 2024 survey.
As the proportion of academic misconduct involving newly generated rather than copied text increases, similarity detection will capture a smaller proportion of the behaviours institutions are concerned about.
Detection of AI generated text is also really difficult. There are AI detection tools but they aren’t well adopted in UK HE. The Pardee Legal Research Center gives a helpful summary of research and risks when it comes to AI detection – with the biggest risk, in my opinion at least, being the possibility of false positives, where the AI detector incorrectly flags text as having been written by AI, when it wasn’t. How can a student prove they didn’t use AI to write the text? How can an institution prove a student did use AI to write the text? There are some reasonably obvious signs (ie AI generates fake references or the student submits the assignment with the prompt included) but otherwise, it is very hard to tell. You can walk through a similarity report and look at the evidence (or lack of) for plagiarism. You can present that evidence to the student being accused of plagiarism. You don’t have a similar audit trail for suspected AI generated text. What even counts as misconduct with the use of AI? There is such a spectrum for using it, sitting between fully generating an assessment through AI and not using it at all. The Detailed Guidance on the Durham University’s Academic Misconduct: AI Policy and Guidance provides 17 different ways a student may use AI in an assessment and whether it constitutes academic misconduct or not.
What does all of this mean for HE (not just in the UK)? I can see institutions moving away from similarity detection longer term because plagiarism methods have changed. Perhaps the future of academic integrity is less about detecting what happened after submission and more about designing assessment that gives us confidence in what the student can actually do.
All roads go back to assessment redesign, as scary as that is….