Last week, the US software company Turnitin revealed 53.6% of Australian tertiary education submissions run through its system used some form of AI in the period from October 2025-April 2026. The company, whose plagiarism-detection technology is widely used in universities, also reported 10% of these submissions contained more than 80% AI-written content.
Meanwhile, separate data from AI safety company Anthropic showed Australia now leads in per capita use of its Claude chatbot. Coursework makes up a significant share of that use.
Our 2026 analysis found while most Australian universities have introduced generative AI policies, significant gaps remain in translating these into practice.
So what can unis do to keep up?
How AI is used matters
The headline figure is striking, but it should be interpreted with care. AI detection tools estimate the likelihood that assisted writing is present, not whether a student breached university rules. A high AI score could reflect disclosed use rather than academic misconduct.
The real concern is not whether AI is present in students’ work, but how it is being used and assessed.
Generative AI can be used in many ways to assist in learning, such as brainstorming ideas, refining structure, improving grammar, generating practice questions, or receiving formative feedback. These activities are increasingly recognised by universities as appropriate when used transparently and within institutional guidelines.
Experts say AI can become problematic when it replaces, rather than supports students’ thinking.
Rethinking how units are taught
A big challenge for universities is distinguishing whether students are using AI to help or as a substitute for learning. Unis need to ask whether their current approach to AI is effective in providing students with the knowledge, judgement and skills required for their degree.
One way unis can approach this is by modifying units to focus on the learning process rather than the final assessment.
For example, instead of asking students to submit a single essay at the end of the semester, an educator might ask them to document how they used AI throughout the unit. They might be asked to evaluate the accuracy of AI responses and reflect on why they accepted or rejected particular suggestions. The final grade would therefore recognise not only the finished product, but also the student’s reasoning, judgement and learning process.
But changing assessment alone is unlikely to improve learning outcomes. Students need repeated opportunities to practise the kinds of thinking, judgement and AI-supported decision-making in adapted assessments. This is why highlighting the learning process as part of a unit restructure is important.
Still, this sort of curriculum change takes time. Teaching activities, assessments, learning outcomes and staff development all need to be reviewed and tested before being introduced at scale.
Tools designed for education
The Turnitin report also found educators are increasingly demanding education-specific AI tools to help them navigate AI use in tertiary education.
Generic AI assistants, such as ChatGPT, are designed to answer almost any question. Universities, however, need AI systems designed around learning, assessment and institutional policies.
Used well, education-specific AI tools can support students through structured brainstorming, provide them with tailored feedback, and guide them to relevant course resources. For teachers, education-specific AI tools can allow them to set limits in how AI can be used by students, and to see how students have used AI in assessments.
Many universities now provide institutionally supported AI tools. Other unis are piloting the technology.
The goal is not to replace educators or automate assessment. It is to create AI that supports learning, transparency and academic integrity. If AI becomes an accepted, routine part of university study, these education-specific systems may become as important to learning as the technologies and software platforms universities currently rely on.
We also need to remember, there is not one simple fix around AI. It’s not something that can be solved by a one-off solution. Unis therefore need to treat AI as an ongoing educational and governance challenge. Effective responses require continuous evaluation, adaptation and investment in teaching, assessment and policy as the technology changes.
Meena Jha receives funding from Australasian Council of Deans of ICT Learning and Teaching Grant.
Amara Atif receives funding from Australasian Council of Deans of ICT Learning and Teaching Grant.
