Almost all university students use AI, but one in six avoids it: the problem is no longer access, it's trust

By The Conversation (via University of Wollongong) · November 10, 2026.

AI is now part of everyday university life, but that does not mean students are comfortable with it. That is the conclusion of a 2026 survey of more than 10,000 students at four Australian universities: Monash, Queensland, Deakin and the University of Technology Sydney. It was authored by Kelly Matthews (University of Wollongong) and Tim Fawns (Monash).

Use keeps growing, though slowly. According to the authors, the share of students who never use AI for study fell from 17% to 15% compared with the same survey in 2024. A separate report, the Australian Student Wellbeing Report, puts the share of students who have used AI for assignments and study at 84%, up 10% on the previous year.

The most common uses are practical. Around 70% use it to edit or improve their writing. 64% use it to generate ideas or examples and 60% to learn how to do something. About half ask it for information or feedback on their work.

The motivations have an emotional component. Roughly 55% use it to work faster and get past mental blocks. 41% say it reduces their stress and 32% say they feel more confident with it.

At the same time, there is solid resistance. According to the authors, 15-17% avoid it entirely and many more limit it. Among those who use it, the most cited reason (67%) is concern about its accuracy. 62% say it clashes with their learning values and 57% prefer to do the work themselves. Almost half voice ethical concerns: bias, privacy, academic integrity and environmental impact.

There is also a striking shift in fear of penalties. This year, 56% say they avoid AI for fear of being caught, 7% fewer than in 2024. Policy clarity is improving: the share of students who say they receive enough guidance from their university rose from 36% to 47%. The authors attribute part of the improvement to more uniform rules per institution, rather than leaving each lecturer to set their own.

**Our take.** The usual narrative pits students who "cheat with AI" against universities that ban it. These data tell a different story: a majority that uses it with judgment and a sizeable minority that holds back for legitimate reasons. That 67% of users distrust its accuracy is not a failing on the students' part; it is a sensible reading of current models' limitations.

Most interesting is the 62% who limit AI because of their "learning values". They understand that effort is part of learning and that outsourcing it can hollow it out. It is an intuition institutions should harness, not correct. Designing assessments in which AI is a tool rather than a shortcut is more useful than policing its use.

A note of caution: this is a self-reported survey, at four Australian universities, recruited via email, internal networks and QR codes. It cannot simply be extrapolated to other countries, and what we have is an academic opinion piece and a published "snapshot", not the full study.

In the short term, the transition in education will be uncomfortable: uneven rules, anxiety and gaps between those who get guidance and those who do not. In the long term, the outlook is better. If AI becomes a reliable, personalized tutor available to anyone, education will become one of the most direct paths toward that abundance of knowledge and opportunity. Getting there, students' trust, which is currently conditional, will matter as much as the models' capabilities. And listening to them, as the authors urge, is the first step.