Turnitin says Clarity cuts AI-written student work

Turnitin has published data showing that students in Australia and New Zealand were less likely to submit writing judged to be almost entirely generated by artificial intelligence when they used its Clarity writing platform. Regional figures were also lower than comparable rates reported for the United States and the United Kingdom.

The findings come from Turnitin's latest Learning Integrity Insights Report, which compared submissions written inside and outside Clarity, a writing environment that records drafting activity and can include an AI assistant. In Australia and New Zealand, 6.7% of submissions written outside Clarity received an AI writing score above 80%, compared with 3.96% of submissions written within the platform.

That represents a drop of about 40% in what Turnitin classifies as near-total outsourcing of student writing. The same pattern appeared in other markets, though starting rates were higher.

In the United States, 14.3% of non-Clarity submissions received an AI writing score above 80%, falling to 7.6% for work produced within Clarity. In the United Kingdom, the rate declined from 9.6% outside the platform to 7.24% inside it.

In Australia and New Zealand, the sample covered 956 Clarity submissions and 312,003 non-Clarity submissions. The US sample included 15,299 Clarity submissions and 2.3 million non-Clarity submissions, while the UK data covered 366 Clarity submissions and 268,818 non-Clarity submissions.

Student use

The report also examined how often students used the AI assistant available within Clarity. Across users in the US, UK, and Australia and New Zealand, Turnitin analysed 840,889 prompts and found that most students did not use the assistant at all when completing assignments.

According to the data, 66% of students completed their writing assignments without using the AI assistant. Another 20% used between one and five prompts per assignment, while 14% used six or more.

Turnitin said this pattern was consistent across secondary and higher education, as well as across the three regions covered by the report. The figures suggest that access to an integrated AI tool does not automatically lead to heavy student use.

Use of the tool differed by education level. In secondary education, students were more likely to use AI early in the writing process. The most common preset prompts among secondary students were for brainstorming, used 4,140 times, and drafting an outline, used 3,764 times.

By contrast, students in higher education were more likely to use AI for checking and organising written work. The most-used preset prompt in that setting was feedback against rubrics and assignment organisation, used 1,976 times.

Policy gap

The report also pointed to a gap between institutional policy and classroom practice. Turnitin reviewed more than 19,000 assignments across 957 institutions in secondary and higher education to assess how the AI assistant was configured.

At the institutional level, 50% of institutions kept the AI assistant always enabled, 45% allowed flexible use, and 5% disabled it entirely. At the educator level, however, the distribution was more rigid: 73% of classes had the AI assistant always enabled, 13% had it always disabled, and 14% fell into a middle category.

These figures suggest that many institutions have adopted policies that allow discretion, but day-to-day teaching settings often turn that flexibility into a simpler on-or-off approach. The report argues that this can leave staff with limited room to tailor AI access to specific tasks.

Turnitin framed the issue as one of visibility in the writing process rather than simple detection of AI-generated text. Its data focused on whether students working in a monitored drafting environment appeared less likely to submit work that was mostly produced by AI tools.

"Transparency is key to better learning outcomes," said Annie Chechitelli, Chief Product Officer at Turnitin.

"A transparent writing space, like Turnitin Clarity, gives educators visibility into how a student's work comes together, allowing them to provide more meaningful guidance. When students understand how their work is being evaluated, the anxiety around AI detection can be reduced, along with the temptation to outsource their thinking entirely. The result is a learning environment that protects critical thinking and nurtures learning integrity," Chechitelli said.

The report adds to a growing body of education-sector research on how schools and universities are responding to generative AI in assessment. For institutions in Australia and New Zealand, the figures suggest student behaviour may differ from larger markets, with lower recorded rates of submissions that appear to have been produced mostly by AI, both inside and outside a monitored writing environment.

Turnitin said the institutional analysis covered users and assignments from July 2025 to August 2026, and the prompt analysis covered activity from July 2025 to August 2026 across the US, UK, and Australia and New Zealand.