While some higher education providers appear to be quietly giving students a green light to use AI without restriction, even when this breaches academic integrity rules, there are many institutions taking a more considered approach. A significant number remain committed to protecting academic standards, yet they are increasingly grappling with the accuracy of AI detection tools and the reliability of impact assessments.
This challenge is particularly evident in specialised courses where the use of AI is embedded in the learning process. In these settings, the issue is not whether AI should be used, but how to ensure that its use is appropriate, transparent and academically sound.
A large proportion of institutions currently rely heavily, and sometimes exclusively, on Turnitin to detect and monitor AI usage. Its accessibility and widespread adoption make it an obvious choice, especially given that many institutions already hold licences for similarity checking. The question remains whether it is the ultimate solution.
Turnitin’s recent Learning Integrity Insights Report offers several noteworthy and at times concerning findings. Ninety four percent of students admitted to using generative AI to assist with assessment completion. Between six and seven percent of student papers recorded an AI similarity score of eighty percent or higher. Fewer than half of institutions have a dedicated AI policy. At the same time, both students and instructors often lack a clear understanding of acceptable AI usage, and even among those who believe they do, there is little consensus.
While these findings highlight genuine challenges, there are also reasons for cautious optimism.
Over time, institutions are becoming more adept at designing assessments that cannot be easily completed through simple AI prompts. At the same time, regulators such as Tertiary Education Quality and Standards Agency are pushing providers to develop, implement and enforce robust AI usage policies. Encouragingly, AI literacy among both students and educators continues to improve, with even those slower to adopt new technologies becoming more confident in using these tools.
However, the report also highlights a deeper and more complex issue, which is the need for customisation. For AI to be used effectively in education, tools must be tailored to the specific needs of courses, institutions and instructors. The same applies to detection systems, which need to be calibrated to the particular tasks and contexts they are assessing.
This is where the path forward becomes more complicated. While licences for popular AI tools are generally affordable, they rarely offer the level of customisation required to deliver optimal results. Generic solutions often fall short, particularly in specialised programs where nuanced content and discipline specific knowledge are critical. For example, AI assisted marking tools without adequate customisation can struggle to accurately assess work in niche or proprietary programs.
If institutions are required to adopt multiple tools across different programs or subjects to achieve the necessary level of precision, the associated development and implementation costs could quickly escalate.
The need for customisation is widely recognised by institutions, educators, students, technology providers and regulators alike. It is actively encouraged across the sector. Yet one fundamental question remains unresolved. Who will bear the cost of building and sustaining these tailored AI solutions?
