AI in Education in New Zealand: A Living Whitepaper
Updated for publication on 9 March 2026
Executive Snapshot
New Zealand’s education sector has moved beyond the first-wave debate of “AI as cheating tool” into a more operational phase: schools and tertiary providers are now building policies, redesigning assessment, testing classroom use cases, and expanding AI literacy. The sector is not yet in full-scale transformation, but it is clearly past the experimental fringe. The strongest evidence of adoption sits in teacher workflow support, school-level policy redesign, NZQA assessment infrastructure, and emerging national AI-literacy initiatives. (education.govt.nz)
Three changes stand out since the previous version of this whitepaper. First, the Ministry of Education has updated its school guidance, including explicit principles for using AI in marking and new New Zealand case studies. Second, AI now appears directly in draft curriculum architecture, especially in the Years 9–10 Technology and Computer Science material now out for consultation. Third, the evidence-gathering phase is deepening: as of March 2026, ERO is actively surveying schools on how AI is being used by teachers, leaders, and students. (education.govt.nz)
Overall State of Adoption
Schools: broad experimentation, cautious formalisation
Classroom adoption is real and growing, but still uneven. Official TALIS 2024 findings for New Zealand show that 63% of Year 1–10 teachers had used AI in some form in the prior year, including 69% of Year 7–10 teachers, well above the OECD lower-secondary average of 36%. Among New Zealand teachers who used AI, the most common tasks were generating lesson plans and activities (78%) and learning about or summarising a topic (73%), while only 12% reported using AI to assess student work. (educationcounts.govt.nz)
That pattern matters: AI adoption in schools is strongest where the stakes are lower and teacher oversight is easy to retain. New Zealand schools are using AI mainly to support planning, drafting, feedback, creativity, and student engagement—not to hand over final professional judgement. This aligns closely with Ministry guidance, which states that AI should improve teaching and learning while keeping teachers central and responsible. (educationcounts.govt.nz)
Tertiary: from detection to redesign
In higher education, the center of gravity is shifting away from AI-detection software and toward assessment redesign, AI literacy, and secured tasks. RNZ reported in September 2025 that several universities had stopped endorsing AI-detection tools, with institutions such as Massey instead prioritising secured assessments and broader AI-literacy development. At the same time, the sector remains unsettled: Victoria University of Wellington’s law faculty moved some exams to handwriting in May 2025 because of AI concerns, while Lincoln University required about 120 postgraduate finance students to explain code in person after suspected AI-assisted misconduct. (rnz.co.nz)
The implication is clear: tertiary adoption is advancing, but trust in older integrity controls is declining faster than a sector-wide replacement model has fully matured. Universities are increasingly accepting that AI use must be governed through task design, oral defence, in-person validation, and clearer expectations, rather than through weak forensic detection alone. (rnz.co.nz)
System agencies: the most mature AI deployment
The most operationally mature AI use in New Zealand education is not in classrooms but in system agencies. NZQA reported that in May 2025 it used an AI-powered Automated Text Scoring tool to mark more than 55,000 writing literacy assessments, returning results 3.5 weeks earlier than the year before. NZQA says a 2024 trial on 36,000 writing samples found the tool as reliable as human markers, and more than a third of May 2025 results were double-checked by experienced human markers, with human judgement prevailing where scores differed. (www2.nzqa.govt.nz)
NZQA is also using AI for service delivery. Its chatbot, Awhina, has been helping users since 2023, and a generative AI upgrade began rolling out in April 2025 to improve question handling by searching and ranking relevant website information. NZQA has said its AI tools already support more than 250,000 student and customer interactions each year. (www2.nzqa.govt.nz)
What Has Changed Since the Previous Report
- School guidance is now more explicit. The Ministry’s generative AI page, last updated on 28 October 2025, now includes updated guidance on AI tools for marking, GenAI in NCEA, and downloadable school case studies. It reiterates that schools must have a policy on acceptable GenAI use, and that GenAI is not permitted in NCEA external assessment. (education.govt.nz)
- Curriculum work has advanced into live consultation. Draft Years 0–10 curriculum content is currently out for consultation until 24 April 2026, with draft Technology material explicitly including AI within the Years 9–10 Computer Science strand. (newzealandcurriculum.tahurangi.education.govt.nz)
- National AI literacy initiatives have moved from concept to rollout. Day of AI Aotearoa launched in 2025 through TENZ, EPIT, and NZCER, with a goal of reaching every school, teacher, and learner by the end of Term 1 2026. In parallel, Kōtui Ako’s nationwide AI Digital Literacy programme opened registrations for classes starting 9 March 2026. (dayofai.org)
- Evidence collection is intensifying. In a 3 March 2026 bulletin, the Ministry highlighted ERO’s new “AI in schools” surveys for leaders and teachers, signalling that national monitoring is still catching up with practice. (education.govt.nz)
Policy, Governance, and Curriculum
Ministry of Education guidance
The Ministry’s position is now clearer than in the initial 2023 response. Current guidance emphasises: do not enter personal data into AI tools; verify outputs; expect cultural bias; keep human teachers central; and ensure schools have a policy covering purpose, scope, privacy, attribution, and acceptable use. The Ministry also notes age restrictions on some tools and explicitly says that use of AI in education should support, not replace, teacher judgement. (education.govt.nz)
On marking, the Ministry’s updated approach is principles-based: teacher responsibility and professional judgement, human oversight, and appropriate use cases. Its accompanying marking guidance warns that AI use for complex work, summative assessments, and credential-related tasks requires extreme caution. For NCEA internal assessment, use of AI to support marking is discouraged because the assessor’s judgement has legal standing. (education.govt.nz)
NZQA and academic integrity
NZQA’s school-facing position remains restrictive for high-stakes learner-authored work: GenAI is not permitted in NCEA external assessment, and schools are expected to manage authenticity through policy and practice. In tertiary education, NZQA advises providers to maintain coherent academic-integrity systems, explain learner responsibilities, use varied assessment methods, support staff understanding of generative AI, and reflect Māori data sovereignty in policy and research practice. (education.govt.nz)
Privacy and Māori data sovereignty
Privacy and sovereignty are no longer peripheral concerns; they are becoming central tests of AI readiness. The Office of the Privacy Commissioner’s expectations for generative AI call for senior approval, necessity and proportionality checks, privacy impact assessment, transparency, and engagement with Māori on risks and impacts. Ministry guidance for schools points directly to this privacy framework. (privacy.org.nz)
The curriculum-facing conversation is also evolving. Tāhūrangi’s AI introduction page explicitly states that Māori data sovereignty is a significant consideration under Te Tiriti o Waitangi and that AI cannot replace engagement with local whānau, hapū, and iwi. This is increasingly shaping what “responsible AI in education” means in Aotearoa, especially compared with more generic international models. (newzealandcurriculum.tahurangi.education.govt.nz)
Curriculum direction
AI now has clearer curriculum visibility than it did a year ago. In the draft Years 0–10 Technology learning area, the Years 9–10 Computer Science strand explicitly focuses on “algorithms, data, and logic in digital systems and AI” and includes the ethical implications of intelligent technologies. More broadly, consultation on the draft national curriculum is open until 24 April 2026, with final release targeted for mid-2026. (newzealandcurriculum.tahurangi.education.govt.nz)
At senior secondary level, the picture is more provisional. Tāhūrangi states that draft Years 11–13 subject curricula are being released in Term 1 2026 for feedback and familiarisation, with no formal requirement to use the new content until 2028 for Year 11, 2029 for Year 12, and 2030 for Year 13. A proposed specialist Year 13 subject on generative AI was reported in September 2025 as under investigation, but it is not yet a settled feature of the curriculum. (tahurangi.education.govt.nz)
Research Overview
Official system evidence
TALIS 2024 is the strongest current national benchmark. It shows New Zealand teachers are ahead of many peers in AI use, but also highly alert to risks. New Zealand Year 7–10 teachers were more likely than the OECD average to believe AI can help improve lesson plans and adapt materials for different student abilities, but they were also more likely to worry that AI lets students misrepresent others’ work as their own and can make incorrect recommendations. (educationcounts.govt.nz)
NZCER evidence from primary schools
NZCER’s 2025 report on primary schools gives a more granular picture of early adoption. Among the surveyed teacher cohort, 18% said they used AI daily and 26% several times a week. The most common teacher uses were lesson planning (82%), assessment design (66%), personalising learning (65%), research for teaching (59%), and writing student feedback reports (51%). At the same time, 68% had observed inaccuracies in AI-generated content. (nzcer.org.nz)
The same report also shows how incomplete current governance remains. Among teachers whose schools had, or were developing, an AI policy, most were unsure whether Māori data sovereignty had been considered; 62% selected “I don’t know.” NZCER also notes that many educators were relying on free public tools, which it warns are often less capable and more prone to error and bias. (nzcer.org.nz)
Student use is also already part of the picture. NZCER reports that self-reported student use was higher outside school than inside, and that teachers estimated at least some student use at home in most cases. This suggests that AI adoption is not something schools can choose to “wait out”; learner exposure is already happening beyond the classroom boundary. (nzcer.org.nz)
Case Studies
1) Aotea College: policy-led secondary response
Aotea College built a staff AI group to create a cohesive school-wide approach. Its work included updated authenticity statements, staff PLD, a handbook covering referencing and verbal questioning, information sessions for students, and assessment guidelines for referencing GenAI use. Teachers were encouraged to add checkpoints, supervise more stages of assessment, know each student’s style, and use verbal questioning where concerns arose. (web-assets.education.govt.nz)
Why it matters: This is a strong example of New Zealand schools moving from ad hoc reactions to institution-level process design. The model treats AI as a governance and pedagogy problem, not just a discipline issue. (web-assets.education.govt.nz)
2) Hobsonville Point Secondary School: authenticity through checkpoints
Hobsonville Point Secondary School responded to plagiarism linked to GenAI by updating policy and requiring checkpoints, standard assessment templates, original working documents with full version history, and verbal checks of understanding. The school paired these controls with restorative practice and family engagement, and the Ministry case study reports a significant drop in inappropriate AI use in assessments. (web-assets.education.govt.nz)
Why it matters: Hobsonville demonstrates a practical middle path between permissiveness and blanket bans. Its use of process evidence, version history, and dialogue is likely more durable than detector-led enforcement. (web-assets.education.govt.nz)
3) New Windsor School and the Ako Hiko cluster: primary experimentation
Education Gazette profiled New Windsor School’s AI pilot as part of the Ako Hiko Cluster. The programme integrates AI into a curriculum that emphasises cyber-safety, creativity, and critical engagement; students use tools such as Brisk for real-time writing feedback, while staff have received PLD support and NZQA AI micro-credential training. The school is also exploring AI for school administration and wider cluster rollout. (gazette.education.govt.nz)
Why it matters: This is one of the clearest primary-sector examples of AI moving from isolated teacher use into cluster-based capability building. (gazette.education.govt.nz)
4) NZQA: AI in high-stakes assessment operations
NZQA’s automated scoring deployment is the most consequential AI adoption in the sector because it has already been used at scale in a national assessment context. NZQA frames this as a quality, consistency, and timeliness tool rather than a cost-cutting exercise, and says wider rollout will occur only where AI proves at least as good as human performance. (www2.nzqa.govt.nz)
Why it matters: Unlike many school-level experiments, this is not exploratory theatre. It is production deployment in a tightly controlled, nationally significant workflow. (www2.nzqa.govt.nz)
5) University of Auckland: service AI and language learning AI
The University of Auckland’s AI Assistant is now handling more than 9,000 conversations and 60,000 searches each month, offering 24/7 support across fragmented university information sources. Separately, the University received a $1 million grant for a three-year project to develop an AI-powered te reo Māori pronunciation coaching tool that will deliver real-time personalised feedback, in partnership with Te Hiku Media and grounded in tikanga Māori and Indigenous data sovereignty. (auckland.ac.nz)
Why it matters: These examples show tertiary AI adoption splitting into two tracks: operational service automation and culturally grounded learning innovation. The second is especially important in the New Zealand context because it links AI capability with language revitalisation rather than only productivity. (auckland.ac.nz)
Emerging Trends
1) AI adoption is strongest in augmentation, not substitution
Across schools and tertiary settings, the winning uses of AI are those that save time, support drafting, or extend access—lesson planning, summarisation, feedback, tutoring, helpdesk support, and literacy development. Full delegation of grading or judgement remains limited and contested. (educationcounts.govt.nz)
2) Assessment redesign is replacing detection as the main integrity strategy
New Zealand’s most credible responses to GenAI misuse are moving toward checkpoints, oral defence, version history, secured tasks, and in-person validation. Schools such as Hobsonville and Aotea, along with university shifts away from AI detectors, all point in the same direction. (web-assets.education.govt.nz)
3) AI literacy is becoming a system priority
Day of AI Aotearoa, Kōtui Ako’s AI Digital Literacy programme, Tāhūrangi’s AI introduction resources, and live curriculum consultation all indicate that New Zealand is starting to treat AI literacy as a mainstream educational capability rather than an optional tech niche. (dayofai.org)
4) Equity, culture, and sovereignty are competitive differentiators
The New Zealand conversation is increasingly distinct because it is tying AI adoption to Te Tiriti, Māori data sovereignty, te reo Māori revitalisation, and culturally responsive design. However, the research shows practice is still lagging aspiration: many teachers remain unsure whether these principles have actually been embedded in school policy. (newzealandcurriculum.tahurangi.education.govt.nz)
5) National policy is still catching up with practice
The sector has active guidance, live pilots, and some scaled deployments, but not yet a fully settled national operating model. ERO’s current AI-in-schools survey work, the open curriculum consultation through April 2026, and the still-provisional status of a senior AI subject all show a system that is advancing quickly but remains in design mode. (education.govt.nz)
Conclusion
As of 9 March 2026, AI adoption in education in New Zealand is best described as active, uneven, and maturing. It is active because teacher use is widespread, NZQA has operational AI systems in production, and AI literacy initiatives are scaling nationally. It is uneven because classroom practice varies sharply by teacher confidence, school policy, and educational level. And it is maturing because the sector is shifting from novelty and panic toward governance, pedagogy, and infrastructure. (educationcounts.govt.nz)
The strategic direction is now clearer than it was a year ago. New Zealand is not moving toward “AI replacing education professionals.” It is moving toward a model where AI supports planning, tutoring, service delivery, and selected assessment processes—while human judgement, cultural legitimacy, privacy, and authenticity become more explicit design requirements. The schools and institutions making the fastest progress are those treating AI as a whole-system change issue, not merely a tool choice. (education.govt.nz)
The defining question for the next phase is no longer whether AI has entered New Zealand education. It has. The real question is whether the sector can scale its use in ways that are educationally credible, culturally grounded, and operationally trustworthy. Current evidence suggests that New Zealand has begun that transition—but has not yet completed it. (education.govt.nz)