In a 1984 essay, Dutch computer scientist Edsger Dijkstra wrote: “The question of whether a computer can think is no more interesting than the question of whether a submarine can swim.”
When we talk about computers “thinking”, artificial intelligence (AI) quickly comes to mind. After all, headlines about “rogue AI agents” allegedly defying their programmers and hacking websites of their own volition imply that this technology is so “intelligent” that it has learnt how to surpass our ability to control it.
To U.S. President Donald Trump, using the term AI “makes the technology sound fake”, hence his official rebranding of AI as “Super Intelligence” (SI) at the United Nations General Assembly on 22 September 2026.
Trump's announcement followed an equally curious statement from OpenAI CEO Sam Altman at the G20 Innovation Ministerial in Chapel Hill on 2 September.
“A kid growing up today will never be smarter than AI,” Altman said.
Any teacher reading that might wonder what that means for their own profession down the line. That said, a growing body of research suggests that AI is far less advanced than what we're being told.
What impact is AI having on learning?
With hundreds of billions of dollars being invested in edtech marketed as “AI-powered”, educators have been keen to understand the impact this technology is having on student learning.
The problem is, the research required to understand this is sorely lacking in Australia.
A major global review by the University of Sydney and the Barker Institute, which spanned 271 studies from more than 45 countries, found that Australia has “little local evidence to guide decisions about generative artificial intelligence (GenAI) in schools”.
What the researchers did find was that while affective gains are common, they do not necessarily indicate learning.
“GenAI use is often associated with increased interest, engagement, motivation, domain-specific self-efficacy, as well as reduced learning-related anxiety. However, it is not clear whether these affective gains translate into better learning outcomes,” the report’s authors said.
There are also concerns over a growing body of research suggesting that heavy reliance on chatbots is silently rewiring young brains for the worse.
A 2025 study by researchers from Central Queensland University (CQU) found that over-reliance on AI dialogue systems hinders vital cognitive skill development.
Students using AI to generate essays showed a 47% drop in active neural connections – from an average of 79 to just 42. More worryingly, a staggering 83% of those students could not recall the content they had written just minutes later, a phenomenon researchers describe as “cognitive debt” or “digital amnesia”.
“Neuroplasticity – the brain’s ability to adapt – is strongest during childhood and adolescence,” Professor Ken Purnell, a leading expert in educational neuroscience, said in a statement. “But that same plasticity means young brains may be rewired by over-reliance on AI, potentially compromising executive function, self-motivation and decision-making skills.”
Other research, including a study by the Australian Network for Quality Digital Education (ANQDE), warns that unstructured AI use risks another malaise – “cognitive atrophy” – in young people.
A more recent study of 26,000 students in China found that while using AI helped kids improve their homework scores by 18% and cut completion times by 30%, it also lowered monthly exam scores by 20% within six months.
Dr Jill McGuire, Group Head of Learning Innovation at Education Perfect, said it’s how the technology is being used, not the technology itself, that holds the key to unlocking its real value in the classroom.
“One of the hardest things for a teacher to do consistently is give every student personalised and actionable feedback while the student still remembers how they arrived at an answer,” Dr McGuire told The Educator. “That is where I see AI’s greatest potential.”
St Mary MacKillop College in Canberra is one of hundreds of schools across Australia using Education Perfect’s AI Feedback Tool on tasks they design and assess against curriculum-aligned criteria.
“Students write their own responses, receive instant feedback, and return to revise them,” Dr McGuire explained. “The teacher can see both the work and the patterns emerging across the class, then decide who needs further instruction.”
An ethics-approved study by Educate Ventures Research (EVR), commissioned by Education Perfect, examined how the platform was being used at St Mary MacKillop College.
One English teacher tracked her students’ drafts and showed the class, anonymously, that every student had improved. Many kept going back to strengthen their work. All nine teachers surveyed also said feedback arrived sooner, giving students a better chance to act on it while the task was still fresh.
“For educators wondering where to begin, I would choose one task where delayed feedback is limiting progress,” Dr McGuire said.
“The College’s study offers a concrete example of visible progress in writing and a more responsive feedback cycle; it does not establish how much of that progress was caused by AI alone.”
While educational AI might be having some cut-through in certain schools, international testing data suggests schools may need to re-evaluate the role of this technology in their classrooms.
The latest results from the Programme for International Student Assessment (PISA) 2025, which involved 760,000 students across 91 countries and economies, found that students who used AI to draft text for writing assignments underperformed compared with those who didn’t.
Responding to the findings, Federal Education Minister Jason Clare called for a “digital detox”.
“PISA data shows the more kids are online, the more their results decline. Other tests like TIMSS, PIRLS and previous PISA rounds show the same thing – students who use screens a lot score lower than those who don’t,” Clare said.
Meanwhile, the NSW Education Standards Authority (NESA) is launching a review to protect student learning and maintain the integrity of the HSC amid the growing use of AI in schools.
The review, ordered by NSW Education Minister Prue Car in August, will probe AI’s toll on students’ cognitive development and push for new ways to ensure major projects and creative work remain students’ own.
“We need to act now as the evidence continues to emerge,” Car said in a statement. “It is better to put sensible safeguards in place while we fully understand AI’s long-term impact on student learning.”
The nation’s online safety regulator agrees.
In August, a report from the eSafety Commissioner stressed the need to “mitigate the risks associated with children using AI assistants and companions, and to carefully investigate how the benefits of these tools might be realised safely”.
The survey of 1,950 Australian children aged 10–17, titled ‘Talking to machines: Children’s experiences with AI assistants and companions’, found nearly 80% are actively using AI.
“A significant minority of the children in our study who had used AI assistants or companions had experienced potentially inappropriate or harmful interactions with these tools,” the report said.
“This includes being given harmful advice, being subjected to controlling or bullying-type behaviours, and being exposed to nude or sexualised images or chats, and to hateful ideas such as misogyny and racism.”
The report went on to caution that many children have also shared personal or sensitive data with AI tools without fully understanding how it might be used, stored or disclosed, pointing to a lack of guardrails protecting young users.
Clare said the sheer volume of screen time in schools is raising some uncomfortable questions.
“When I was a kid there was a computer room at school. Now they are everywhere, from morning till night,” Clare told The Educator. “But they’re not always being used for learning, and AI is supercharging this.”
Clare said a line needs to be drawn between using AI and outsourcing learning to it.
“We know a lot of teachers and parents are worried that students are using AI to do the thinking for them,” he said. “That’s not learning.”
Lessons from the world’s AI superpower
A recent survey by the American College of Education found more than half (54%) of U.S. teachers have received no training whatsoever on using or managing AI at school, even as the technology’s footprint in their daily work looks set to grow.
Dr Jodi Feikema, Interim CEO and Provost, American College of Education, was asked if introducing AI into schools before the research had properly evaluated it might have been premature.
“I think schools have moved faster than the evidence in some areas, but it’s also important to remember that generative AI is likely already part of students’ lives outside of the classroom,” Dr Feikema told The Educator. “It’s true that we still don’t have a settled picture of what regular use does to memory, reasoning, and problem-solving over time, but it also feels naive for schools to ignore it entirely.”
If anything, said Dr Feikema, students could benefit from a teacher’s guidance on how to use AI effectively and, most importantly, how to use critical thinking to assess the accuracy and credibility of what AI produces.
“My concern is less about students getting an answer from AI and more about them skipping the mental work that helps learning stick, but I wouldn’t respond with a blanket ban,” she said. “School leaders need to clearly state which tasks students need to be able to do unaided by AI.”
Dr Feikema said students should still have to show that they can reason, write, solve, and explain without an AI tool doing most of the work.
“And they should also show that they have the skills to assess AI output itself, analysing for validity, checking its sources, and recognising when it falls short.”
The survey’s findings follow an announcement by New York City Mayor Zohran Mamdani on 2 September imposing a one-year moratorium on generative AI for elementary and middle school students while the city spends the next year studying the impacts of the technology.
The sweeping ban, which applies to more than 600,000 students, includes all student-facing AI software, under-eights’ personal devices, and companion chatbots for all ages. However, teachers can still use AI for planning – just not for grading student work.
“Children need to develop skills alongside their peers, build relationships with educators and wrestle with tough problems on their own,” Mamdani said in a statement. “The tech industry wants us to believe that AI-powered early education is not only inevitable, but necessary. We do not see it that way.”
Clare said he addressed New York City's ban on AI in schools during a parliamentary session a few weeks ago.
“I also spoke about how the evidence shows the more kids are online the more likely they are to fall behind,” Clare said. “I’m not making an argument against technology in schools. Students need to learn digital skills. That includes how to use and not use AI. But I think there is an argument to wind it back a bit.”
Clare said there is also evidence that when children write something down with pen and paper, they absorb it more into their long-term memory.
“That’s why Education Ministers have sought advice on whether next year’s NAPLAN test should be done with pen and paper,” he said. “It’s part of a bigger issue Education Ministers are looking at, and it will be on the agenda when Education Ministers next meet.”
Why other countries are rethinking AI in schools
The European Union (EU) has also proposed a ban restricting under-15s from using AI chatbots in the most sweeping plan yet to protect children from online dangers.
The draft EU Kids Act, if signed into law, would also bar under-15s from social media, video-sharing platforms and online games altogether.
Teenagers aged 15 and over would be free to set up their own accounts, while those aged 13 and 14 could ask a parent to create an introductory account on their behalf, but only under parental oversight, and with limited contact lists and strict time limits attached.
For children aged three to 12, accounts would be fully controlled by parents and restricted to child-friendly services that meet stringent safety standards, according to the proposal. Children under three would have no access to the platforms at all.
The European Commission document, seen by Reuters, said the EU’s approach should allow minors to benefit from the huge potential of digital services, while protecting them from abuse and exploitation.
“It needs to limit the access of tech companies to our children, and not the other way around,” the document said.
The EU’s proposal follows Norway’s crackdown on AI, which will see students aged 6–13 barred from accessing generative AI tools in the classroom as a general rule, while those in lower secondary school, aged 14–16, can cautiously use the technology so long as it’s under teachers’ supervision.
Earlier this year, the UAE Ministry of Education released the “Safe and Responsible Use of Artificial Intelligence in Classrooms 2026” guide to ensure early education is built on human interaction, tactile creativity, and independent problem-solving.
The framework, announced in February, sets out 25 strict rules and prohibitions for generative AI use in schools and restricts children aged under 13 from using it in the classroom and during exams.
When asked if schools should be hitting pause on the use of AI in classrooms until this technology and its impact are better understood by educators and education policymakers, Professor Rose Luckin CBE – Professor Emerita at UCL and founder of Educate Ventures Research, which was commissioned by Education Perfect to carry out its impact research – argued that the real risk lies not in using AI, but in failing to teach students how to use it well.
“A blanket pause treats very different tools as one thing. A practice tool that asks a student to explain each step is not the same as a chatbot that writes an essay,” Professor Luckin told The Educator.
Citing the University of Sydney and Barker Institute review, Professor Luckin pointed out that “AI helped least when it did the thinking students were meant to develop”.
“In PISA 2025, among students who use AI regularly to help them learn, those taught to judge the quality of AI output tend to score higher in science, and disadvantaged students are less likely to be taught this,” she said.
“A pause removes the chance to teach it. The Minister is right that children need to learn how to use AI and how not to. My advice is to learn fast and act more slowly.”
Andy Mison is the President of the Australian Secondary Principals' Association (ASPA), the peak national body that represents and supports more than 2,200 public combined and secondary school principals across Australia.
Mison believes it’s timely for Australia to consider some regulation around the use of AI.
“[Regulation] can help us understand what the research is telling us about it before we rush headlong into a potentially dangerous AI universe,” Mison told The Educator.
“I’m mindful about how we launched into smartphone use around 2012–2013 before properly understanding its impacts on young people and their learning, and we’ve been seeing the negative results of that.”
Mison pointed out that Australia has only recently put regulations in place regarding what is safe for kids to have access to online.
“My colleagues and I have welcomed the social media ban, which we believe is a step in the right direction,” he said. “I think it’s now a good time to review how we’re using AI in education, but let’s not impose knee-jerk blanket rules.”
Mison said if restrictions are introduced, they must be introduced in consultation and collaboration with school leaders and teachers.
“Most teachers are using tech in meaningful ways, and we don’t want any unintended consequences of rashly throwing the baby out with the bathwater,” he said.
“Let’s slow down, and let education be part of [the] broader conversation and debate around the role that AI and technology more broadly has in our schools; but let’s also move through that based on the evidence, sensibly and methodically.”
Mison added that this should not be “yet another thing that schools have to manage without being consulted and supported”.
Workload pressure keeps AI on the agenda
New data suggests that the use of AI in Australian classrooms is growing in tandem with the advancements being seen with this technology.
The EdTech Adoption in Australian Schools 2026 Research Report, released this week by Campion, found 56% of respondents planned to expand their use of education technology.
General-purpose generative AI is also becoming part of teachers’ planning and preparation, used by 45% of respondents for planning and 55% for lesson preparation. Teaching and learning tools are most prominent in student practice and application, where they are used by 50%, and classroom teaching, at 46%.
“I don’t think schools have necessarily put the cart before the horse [with AI],” Campion Group’s General Manager of Growth, Brett Watson, told The Educator, adding that this technology is evolving quickly, and the opportunity is “significant for both teachers and students”.
“For teachers, it can reduce workload across planning, resource creation, assessment and feedback, creating more time for teaching and student support. For students, well-designed AI can provide practice, feedback, scaffolding and more personalised learning experiences.”
The key, said Watson, is purposeful use. He pointed to Campion’s research, which reveals strong interest in education-specific generative AI. According to the company’s survey, 60% of respondents reported being ‘very’ or ‘extremely’ interested in expanding their use of AI, while 94% identified time and workload as a major barrier.
“That tells us schools are looking for technology that genuinely improves learning and reduces friction,” he said. “The sensible approach is to pilot carefully, align AI to clear educational goals and ensure it complements strong teaching – and physical resource provisioning – rather than simply adding another layer of technology.”
Whichever way the struggle between the ‘more tech’ and ‘less tech’ camps pans out in education, the Australian economy’s appetite for AI, broadly speaking, remains strong.
Between 2015 and 2024, Australian firms filed more than 600 AI-related patents, while business investment in AI R&D more than doubled, from $276.3m in 2021–22 to $668.3m in 2023–24, according to IP Australia and the ABS.
Meanwhile, Westpac IQ’s ‘Powering the AI Economy’ report shows Australia spent $8.7bn on data centres in Q1 2026 alone – nearly 20% of all private capital spending nationally.
So, there’s no question that wherever there is AI, big money isn’t far behind.
Research firm Gartner forecasts global AI spending will hit US$2.59tn by the end of 2026 – a 47% year-on-year increase – as businesses around the world adopt this technology at an increasingly rapid rate.
The ROI question
For all its hype, questions persist about AI’s return on investment – not just for student outcomes, but in real dollar terms – casting doubt on the sustainability of the business model driving the global AI boom.
PwC’s 29th Global CEO Survey, which included responses from more than 4,400 CEOs in 95 countries, found three in 10 reported an increase in revenue from AI products, and roughly a quarter reported lower costs, while just one in eight were seeing both.
The survey’s findings follow a separate study by MIT, which found most organisations surveyed aren’t seeing bang for buck on their AI pilots. Titled 'The GenAI Divide: State of AI in Business 2025', the study found that despite US$30–40bn in enterprise investment into GenAI, 95% of organisations are getting zero return.
Dr Clare Southerton, a senior lecturer in Monash University’s School of Education, Culture and Society, said she would not be surprised if AI turns out to deliver very little return when it comes to student learning.
“While I can see that there are potential use cases of AI in education, I think these are far fewer than initially imagined and the benefits of these few use cases do not outweigh the negative cost of GenAI on learning that we are already seeing,” Dr Southerton told The Educator.
“That’s not even starting to consider the environmental costs and the ethical concerns with most commercial models, which are largely built on stolen data and exploited labour.”
But beyond the question of whether AI in Australia’s classrooms is getting a good return on investment, Dr Southerton said we have to ask at a basic level: even if it were producing good results, is this the kind of education system we want to foster?
“Do we want to foster an education system where learners become highly dependent on such a destructive and unethical technology?” she said.
“To me it seems like the opposite of the vision of education, which is to provide students with capacity, with the skills to think for themselves and to go out into the world equipped for change.”
Schools must teach about AI, not just with AI
While education ministers, ‘techsperts’ and academics concede that “AI is here to stay”, what is less clear is what role it will play in schools as its risks become more potent.
“We certainly need more Australian-based research to understand how AI use is impacting learning,” Dr Southerton told The Educator. “We do think the overall enthusiasm for AI tools needs to be tempered in light of this [eSafety and PISA] evidence.”
Dr Southerton said communities also need to be mindful of the big commercial interests pushing many of the AI products marketed to schools.
“We need to be realistic about the fact that most of the tools students are using are commercial products designed to maximise engagement,” she said. “In essence, building dependence is good business; they’re not designed to maximise learning.”
Dr Southerton said building AI literacy is key in this regard.
“I don’t think that necessarily means teaching with AI but rather can be teaching about AI,” she said. “There is a common narrative that students will miss out on developing skills to use GenAI if they don’t get the chance to access it, but I think we need to balance this with the development of other essential skills like reading and writing.”
A matter of governance and judgement
From where Andy Mison sits, AI use in Australian classrooms is far from rampant. A more accurate picture, said Mison, is a school system that has become increasingly pragmatic and cautious about AI.
“There hasn’t been an all-in, reckless approach to AI from schools. Having said that, there’s certainly been some engagement with it,” Mison said. “For example, teachers and leaders have been using it to ease workloads, such as some low-level admin tasks, but I think people now have a much more realistic understanding of what AI can and can’t do.”
Mison said most teachers in most schools are “probably at the point where they now recognise that AI is not the silver bullet it was initially thought to be”.
“Like any other tech innovation, it requires humans to exercise judgement,” Mison said, adding that he is increasingly concerned about Big Tech’s statements on AI.
“We do need some guardrails around AI, especially in light of the hack against Medicare,” he said.
“Let’s recognise that it may have some potential benefits for education, but [it] must be carefully tested, and then mediated through teachers’ professional judgement.”
EthosOne is governance, risk and compliance software built only for independent schools in Australia and New Zealand. The organisation’s co-founder, Dave Yeates, disagreed that the cart has been put before the horse with AI in schools, but acknowledged the need to walk this line carefully.
“The risks are real, and over-reliance is one of them, but in many ways, the horse has already bolted, so this is now a case for steering, not stopping,” Yeates told The Educator. “We need to understand where AI can cause harm, get much better at managing those risks, and also identify where it can quietly do good.”
Yeates said “technology has always changed what is asked of us”, adding that calculators were accused of dulling minds, yet made advanced mathematics more accessible.
“AI presents a similar challenge,” he said. “Used poorly, it can create new risks. Used well, it can reduce cognitive overwhelm, support executive functioning in some contexts, and dramatically improve accessibility and differentiation across different abilities.”
Yeates said the task for schools is not to choose between embracing AI and rejecting it.
“It’s to build the governance and judgement to use it responsibly.”
What happens when the AI is taken away?
Dr Charlotte Pierce is a senior lecturer in Monash University’s Department of Human-Centred Computing. She said the AI push into schools was premature and that education systems need to double down on human-centred approaches.
“I don’t think we’re being nearly cautious enough about this,” Dr Pierce told The Educator.
“It’s been in the news recently that Gen Z is the first group in U.S. history to perform worse on standardised testing than previous generations, and many experts attribute this to the increased use of technology in the classroom.”
Dr Pierce said these findings relate only to how technology in general is distracting students and diverting their attention.
“The potential for AI to have a negative impact is even larger,” she said. “It’s addictive, and there are enough concerns around how it affects children that I think we need to be much more careful.”
Dr Pierce said building students’ AI literacy is going to be important in the same way that teaching online safety became important once the internet became widely available.
“I also think it’s absolutely critical to take an approach which centres humans,” she said. “So much about learning happens in those ‘in-between’ moments, where a student explains what they’ve done, or when they overhear someone else getting help, or when they navigate a disagreement with a peer.”
Dr Pierce said these are not things AI can replace.
“I also can’t overstate the value of personal connection from a learning coach like a teacher in motivating and supporting students.”
Professor Luckin sees big potential for AI to improve learning, especially at the moment a child gets stuck.
“The children who most need help are often the least likely to ask for it. AI can, if it offers a hint rather than the answer,” she said.
Professor Luckin said a major study out of Turkey shows why design matters.
Working with nearly 1,000 secondary maths students, Hamsa Bastani and her colleagues compared three groups. One practised with a standard chatbot, another with a version designed to give hints, and a third without AI at all.
“Tested later without AI, the standard chatbot group scored 17% lower than students who never had it, while the hint-based design largely removed that harm,” Professor Luckin said.
“So, ask one question of any AI tool: what can a child do once it is taken away?”
Where to from here?
Brett Salakas, who has worked in Government, Catholic and Independent schools for 28 years, understands the challenges associated with this shift better than most.
After finishing up his role as a consultant with the NSW Department of Education in April 2021, Salakas became HP Australia's education ambassador, where he helps schools meaningfully integrate technology into their classrooms.
He said that, in the short term, the benefit of AI is very squarely on the teacher.
“AI in education is something that should be looked at to relieve time from the teacher so they can better spend their time on being with the student,” Salakas told The Educator, adding that, if handled well, AI could bring teachers and students closer together rather than pushing them apart.
“I believe if we get AI right, right policies, guidance from our institutions and right political vision, it will allow teaching and learning to be more human than ever before,” he said.
“In the immediate term, AI is doing automated admin tasks, and in the mid-term, I think it will likely be used more for research purposes.”
Salakas said his hope is that, in the long term, teachers who use AI will revive the ancient Greek understanding of the paidagōgos (“child-leader”) – a trusted guide who walked a child to and from school, carried their books and watched over their behaviour and development.
“I can see teachers becoming more humanistic in helping kids learn, because education is really about human development and helping the child grow holistically.”
Salakas said there also needs to be more understanding around how children can learn with AI aids, especially as the technology becomes more advanced.
“We know there is risk, so we need to do it right,” he said. “The most effective part is not only to look at how we can use AI, but how we should assess learning.”
Salakas said there is a trove of research showing what good quality assessment looks like.
“We don’t want that low level stuff. We have different hierarchies of knowledge, and we know where assessment should be, but for whatever reason, we often do those low-level information tasks,” he said.
“So, I really think the conversation on AI needs to be coupled with what effective assessment is, because if we get assessment right, concerns around AI become diminished.”
Salakas said teachers ultimately need to leverage devices to decrease the volume of administrative work and spend more time with students “with good practices we know work”.
“Maybe AI gives us the impetus to stop rolling out simple, ineffective assessment strategies and forces us to move into known, well-researched, best practices that challenge higher-order thinking,” he said.
“There are potential risks revolving around poor choices that can be made as this technology advances, but I ultimately hope we have bold, brave leadership to move in the right direction.”
Mison said Federal Education Minister Jason Clare and the Commonwealth Department of Education are “actively engaged” in the conversation about AI in schools.
“I think this government is responding to this [AI] issue in a very sensible, professional way,” Mison said.
“If you want effective policy in schools, it makes sense to engage with those who are implementing it, not just at the end but through the design stage, from the outset, and we’ve got a constructive dialogue with the Federal Government underway on these issues.”
Mison said there is now broad recognition that “the benefits must outweigh the risks when it comes to using AI in schools”.
“If the benefits are there, we need to mitigate against the risks, but in a sensible, measured way,” Mison said. “If we do decide there is a place for AI in education, we need to put guardrails around it.”
Mison said he is particularly glad that this discussion is being had early, noting that the conversation about smartphone guidelines and restrictions came long after they were introduced into classrooms around the country.
“Australians have been very early adopters when a new technology is introduced, but I think we can all now see the need for a more careful response to AI.”