Most teachers think AI is already blunting the way students think. Among secondary teachers, it’s closer to three quarters.
Of the teachers surveyed for a new Education Review Office review, 57% said AI has had a negative effect on students’ critical thinking, rising to 72% at secondary level. Students see it as well. One told researchers that AI gives you the answer but does not require any thinking of your own.
A small group of teachers has opted out on principle, citing the example it sets for students and the effect on creativity and independent thought. For everyone else, the question is no longer whether students use AI, but what they give up when they do.
The uptake is close to total. ERO found more than nine in ten school leaders and more than four in five teachers now use AI regularly, mostly to cut administrative work, draft communications, plan lessons and tailor resources.
Three in four students use it too, mainly to find information and generate ideas. There was no meaningful digital divide by decile or socio-economic background, a result Chief Review Officer David Ferguson said surprised the review team, given how unevenly other technologies have reached schools.
The habit is also hard to break. Around 38% of secondary students and 20% of primary and intermediate students admitted using AI in ways they knew they should not, usually to get through repetitive tasks, work they felt they already understood, or assignments competing for time with other subjects. As one primary student put it, once you start using it, it becomes difficult to stop.
Assessment looks better. Just 14% of students reported using AI inappropriately for assessed work, and almost three quarters said they never had. Secondary students were the exception, with 19% reporting inappropriate use and a further 9% unsure whether their use crossed a line. Only 5% of secondary students and 7% of primary students said a teacher had caught them.
Most students expect AI to be off limits in assessment unless a teacher allows it, even where nothing is written down. Little is. 73% of teachers had never seen the Ministry’s existing guidance on generative AI, and most schools have no school wide policy, which leaves individual teachers to make their own calls.
Ferguson said the answer is not to pull back, given how embedded AI already is in school life and in the world students are being prepared for. ERO wants a national AI in education framework and clearer national policies on teacher and student use. Some schools are already there, with internal policies, designated AI champions on staff and expectations set with students. ERO holds them up as models for a more strategic approach nationally. The review, Ready or not: How are schools responding to Artificial Intelligence?, drew on almost 4,000 survey responses from students, teachers, leaders and parents, along with interviews, focus groups and visits to eight schools.
Beyond the classroom
Separate research released this month found the same gap outside school. Netsafe, the Artificial Intelligence Safety Forum and The Light Project surveyed more than 1,200 New Zealanders aged 13 to 24, with support from the Spark Foundation, and almost nine in ten had used generative AI in the past six months.
Netsafe chief executive Brent Carey says the argument about access is over. “The conversation has moved well beyond whether young people should be using AI. They already are. The challenge now is making sure the adults around them understand the technology well enough to help them use it safely, critically and confidently.”
Of those surveyed, 30% had used AI for personal advice or emotional support, often because it felt easier than talking to a person, or because they did not expect to be judged. Many said they worried about relying on it too much, and wanted the adults in their lives to understand what AI can do and where its limits are. Only 31% would go to a parent or caregiver first if something went wrong online, and almost one in four would not know who to ask or would tell nobody. Fear of trouble, embarrassment and a belief that adults do not understand AI were the main barriers.
Nikki Denholm, director of The Light Project, says youth workers see this too.”Young people are curious, adaptable and incredibly capable, but many are navigating technologies that adults don’t yet fully understand. As their guardians, the most protective thing adults can do right now is to build their own knowledge of AI and create positive, informed and shame free conversations where young people feel safe to ask questions, speak openly and seek support.”
The harm is not hypothetical. Nearly seven in ten respondents had met at least one AI related harm, including misinformation, AI generated scams, misleading content, unsafe advice and deepfakes. About one in nine had experienced a deepfake or knew someone who had.
Shannon Novak, director of the Artificial Intelligence Safety Forum, says safeguards are lagging. “AI is evolving faster than many adults, schools and organisations can adapt. The answer isn’t to discourage young people from using it, it’s to equip them with the skills to question what they’re seeing, recognise misinformation, protect their privacy and understand where AI can help and where it can’t. AI literacy is becoming as fundamental as digital literacy.”
Young people were not asking for less technology. More than half put better information for young people first, ahead of stronger privacy protections, better education on deepfakes and scams, and clearer guidance for parents and whānau.
A thinner pipeline
Tech New Zealand warns the country’s digital skills base is shrinking at the same time. Participation in NCEA technology subjects fell about 2% a year between 2015 and 2024, and the share taking science, technology and maths subjects dropped from the high 80% range to below 75%.
Chief executive Graeme Muller says that is not a passing dip. “This is no longer just a leaky pipeline. It is a structural challenge,” Muller says. He wants stronger foundations, clearer pathways and better coordination between industry, government and education, so the country builds digital capability rather than importing it.
His report also flags an “AI paradox”. AI tools help people finish digital tasks faster, but can add complexity, technical debt and security risk when used without enough underlying expertise, which makes digitally literate graduates more valuable, not less.
All of it will get an airing during Netsafety Week, running nationally from 17 to 23 August under the theme “Online Safety is a Team Sport, Play Your Part”. Sessions include a look at AI and online safety in Aotearoa, and registration is free.
Is something being lost? Most teachers believe students’ thinking has already suffered. Students say the shortcut is hard to give up once taken. Nearly a third are taking personal problems to a chatbot, and fewer than a third would take them to a parent. Neither study argues for switching AI off. Both say the adults need to understand it well enough to answer the question.
