Artificial intelligence detection tools can wrongly flag students and miss AI-written work, raising concerns about fairness, stress and how universities assess academic integrity.
Research from James Cook University found detection software was often wrong, claiming both false positives and missing examples of AI-generated work.
The tools could create more harm than good, only adding to an already stressful experience being a student, researcher Wayne Bradshaw said.
"You get people trying to write like they're not AI, which then adds whole layers of stress and existential dread," Dr Bradshaw told AAP.
"(Students think) I have to write in a way that is not natural to me because I'm concerned about sounding like I'm a machine.
"You have people inserting artificiality into their work to make it sound less artificial."
Words such as "delve" or "tapestry" and the use of em dashes are considered telltale signs of AI, but may just be part of a student's vocabulary.
"They're words which were frequently used by people before they were used by AI,'' Dr Bradshaw said.
"And the reason they're used by AI is because they were used by people."
Students with autism, or from linguistically diverse backgrounds, can also be unfairly targeted by AI detection tools.
It comes as NSW Education Minister Prue Carr has called for a ban on take-home assignments for the state's year 12 students, due to the threat of AI.
The Townsville-based researcher understands the sentiment, and believes universities should be redesigning assessments to be more easily supervised, including pen and paper exams, discussions, experiments and fieldwork in the grading process.
However, he says assignments can't be completely abolished.
"We need students who can write longer texts, who can engage in large-scale projects, who can problem-solve in real-world contexts," Dr Bradshaw said.
James Cook University's Deputy Vice Chancellor Mitch Parsell says trying to catch AI use with these detection tools can establish a combative relationship between institutions and students.
It's why the university doesn't use the tools.
"Our approach to securing learning is to focus on the learning design, not on detection," Mr Parsell told AAP.
However, there is still a process if the university suspects a student is passing off AI-generated work as their own.
"It would involve inviting the student in, in a supportive way, to have a conversation around the assessment task, to see if they actually understand the material that they've submitted," the deputy vice chancellor said.
