A major review of academic research into AI and generative AI in higher education has found teaching staff are broadly open to the technology.

Many are being asked to manage one of the biggest shifts in modern education without enough training, policy direction or institutional support.

The findings point to a growing public-interest problem for universities: AI is already reshaping teaching, assessment, student support and academic integrity, but the systems meant to govern it are still catching up.

The review, published in the Australasian Journal of Educational Technology, examined 29 empirical studies published between 2018 and 2023, covering 4,341 university teaching academics across multiple countries and disciplines.

Of those studies, 19 focused on traditional AI and 10 examined generative AI tools such as ChatGPT.

The conclusion is not that academics are rejecting AI. Far from it.

Most studies found university educators held somewhat or largely favourable views towards AI and GenAI, particularly where the tools could reduce repetitive work, support lesson planning, generate teaching materials, personalise learning or provide faster feedback to students.

But the optimism comes with a heavy qualification.

Academics are worried about cheating, unreliable outputs, fabricated information, privacy, bias, student overdependence and the erosion of core skills such as critical thinking and independent writing.

In other words, universities are not dealing with a simple technology adoption issue. They are dealing with a structural education problem.

The AI Ban Days Are Already Over

When ChatGPT burst into public use in late 2022, many universities treated it as an academic integrity emergency.

Some institutions moved quickly to restrict or ban generative AI tools, fearing students would use them to complete essays, exams and assignments with little or no original work.

That fear was not imaginary. The review notes that GenAI’s ability to pass assessments helped trigger serious concern among educators about widespread cheating. But the research also shows that the debate has moved well beyond plagiarism.

Teaching academics are now weighing whether AI can be used to improve education rather than simply undermine it.

The review also found GenAI is being considered for course planning, assignment design, research support, writing assistance, translation, feedback, learning materials and student guidance.

That shift matters because blanket bans are becoming less practical. AI tools are now embedded into search engines, workplace software, writing platforms, coding tools and learning systems.

Students are not stepping into an AI-free labour market after graduation. Universities know it, even if their policies haven’t caught up.

Staff See The Benefits — But Not The Guardrails

One of the clearest findings is that academics see AI’s practical value.

Traditional AI was viewed as useful for administrative and system-level tasks, including decision-making, student tracking, learning analytics and automated support.

GenAI, by contrast, was valued for its ability to create content: ideas, outlines, lesson plans, assessment material, writing prompts, summaries and teaching resources.

The review found customisation and personalisation were among the strongest perceived benefits across both AI and GenAI.

In theory, universities could use these tools to deliver more tailored support to students at scale — something higher education has promised for years but struggled to achieve in practice.

For overloaded academics, that is a serious attraction. AI can help with the routine, repetitive and time-consuming parts of teaching. But the research makes clear that efficiency alone is not enough.

The problem is not whether AI can save time. The problem is whether universities can prove it is being used responsibly, accurately and fairly.

The Hallucination Problem Is Now An Education Problem

The review draws an important distinction between traditional AI and generative AI.

Traditional AI tends to classify, predict or automate based on existing data. GenAI creates new material. That generative power is what makes it useful — and dangerous.

Tools such as ChatGPT can produce fluent, confident and persuasive answers that may be wrong, incomplete or entirely fabricated. In education, that creates a serious risk.

A student may not know when an AI answer is false. A staff member under pressure may not have time to check every claim. A poorly designed assessment may reward polished output rather than genuine understanding.

The review identifies accuracy and reliability as major concerns, especially because GenAI can “hallucinate” false answers.

But it also points to a more constructive possibility: universities could teach students to identify and challenge AI errors as part of learning.

That may become one of the defining assessment shifts of the next decade. The question may no longer be, “Did the student use AI?” It may become, “Can the student evaluate what AI produced?”

The Real Gap Is Training

The most damaging finding for universities is not that academics are sceptical. It is that many are under-supported.

Across the studies reviewed, academics repeatedly reported a lack of formal training, clear policy, institutional guidance and practical support. This was not a minor side issue. It was one of the main barriers to adoption.

The review argues that institutions need comprehensive training programs, dedicated AI support structures, ethical guidelines, pilot programs and stronger policy frameworks.

Without those supports, universities risk leaving individual lecturers to make high-stakes decisions on their own.

That is not sustainable.

A lecturer deciding whether a student’s AI use is acceptable should not have to rely on instinct. A course coordinator redesigning assessments should not be guessing. A faculty adopting AI feedback tools should not be operating without privacy, bias and transparency safeguards.

The Public Interest Stakes Are Larger Than Cheating

The temptation is to frame AI in universities as a student cheating story. That is too narrow.

This is about the credibility of degrees, the future of assessment, the workload of academic staff, the skills students take into the workforce, and whether universities can adapt quickly enough without lowering standards.

The review points to three skills that may become more valuable as AI becomes more common: strong evaluative judgement, the ability to find original solutions, and the ability to use AI efficiently and responsibly.

That is a useful warning. If AI can generate a passable essay, basic summary, code sample or research outline, universities will need to place more value on judgement, verification, originality and applied reasoning.

Universities Need To Move From Reaction To Strategy

The research ultimately presents a picture of cautious optimism.

Academics are not anti-AI. Many can see its value. They believe it can improve productivity, support teaching and personalise learning.

But they are also alert to the risks: academic integrity, unreliable outputs, dehumanised learning, privacy concerns and weakened student capability.

That tension is now the centre of the higher education debate.

Universities can no longer treat AI as a temporary disruption or a misconduct problem to be policed at the edges. It is becoming part of the operating environment of higher education.

The institutions that handle it well will not be the ones that simply buy new tools or issue vague policy statements.

They will be the ones that train staff properly, redesign assessment intelligently, protect academic standards and teach students how to work with AI without surrendering their own judgement.

The research makes the direction clear: AI is already inside the university system. The question now is whether universities are prepared to lead it — or whether they will keep asking academics to improvise while the technology moves ahead without them.