AI in Education in New Zealand: A Living Whitepaper
Updated for publication on 10 June 2026
Executive Snapshot
AI adoption in New Zealand education has advanced further since the last update on 2 April 2026, but the most important shift is not simply “more use.” It is more structure. In schools, the Ministry of Education has now refreshed its guidance to include clearer rules for AI-assisted marking. In tertiary education, major universities are moving from broad principles to operational models for assessment, tool access, and governance. At system level, NZQA remains the clearest example of scaled, production AI in education, while curriculum and qualification reforms are setting the conditions for how AI will be handled in teaching and assessment over the next several years. (education.govt.nz)
The sector is therefore best described, as of 10 June 2026, as broadly adopted, policy-thickening, and still uneven. AI is now mainstream enough that the central question is no longer whether educators and students are using it; the live question is how New Zealand institutions govern it without weakening authenticity, privacy, cultural legitimacy, or teacher judgment. That reading is supported by TALIS 2024, NZCER’s recent primary-school AI research, the Ministry’s updated guidance, and the growing number of tertiary operating models now visible in public documentation. (educationcounts.govt.nz)
Three developments matter most right now:
- School guidance has become more explicit, especially on AI-assisted marking, with the Ministry stating that AI must support rather than replace teacher judgment and that use for NCEA internal assessment marking is discouraged. (education.govt.nz)
- The curriculum and qualification timetable has shifted, with the Government extending much of the Years 0–8 rollout to 2029, while keeping all Years 9–10 content for 2027 and confirming the phased replacement of NCEA from 2028 to 2030. (tahurangi.education.govt.nz)
- Tertiary institutions are now visibly codifying AI use, with the University of Auckland’s Two-Lane assessment model moving toward full implementation by 2027, Cogniti being opened for more 2026 course deployments, Massey maintaining an assessment-level AI permissions framework, and Otago adopting a university-wide AI Governance Policy effective 10 March 2026. (teachwell.auckland.ac.nz)
What Has Changed Since 2 April 2026
- Ministry guidance is now more operational. The Ministry’s Generative AI page was updated on 22 May 2026 and now foregrounds AI tools for marking. The accompanying marking guidance says teachers remain fully responsible for marking decisions, human oversight is essential, students should be told when AI is used in marking, and AI use for marking NCEA internal assessments is discouraged. (education.govt.nz)
- Curriculum timing has been reset. On 14 May 2026, Tāhūrangi confirmed that while Years 9–10 content still must be taught from 2027, most remaining Years 0–8 learning areas now do not become mandatory until 2029. This slows some curriculum-level AI embedding in the junior years while preserving the secondary timetable. (tahurangi.education.govt.nz)
- The post-NCEA system is more concrete. On 16 May 2026, the Government confirmed further details of the qualifications replacing NCEA: a Year 12 New Zealand Certificate of Education (NZCE), a Year 13 New Zealand Advanced Certificate of Education (NZACE), and a Year 11 Foundational Award benchmarked to literacy and numeracy, alongside subject-based assessment and at least one exam per subject. (tahurangi.education.govt.nz)
- AI literacy rollout has gained stronger local evidence. Education Gazette’s 24 April 2026 feature on Day of AI Aotearoa reported that the 2025 pilot covered nine schools, that 92% of students had already tried AI tools at baseline, and that student ability to explain AI rose sharply after the programme in both primary and secondary settings. (gazette.education.govt.nz)
- University operating models have become easier to see. The University of Auckland now states that all courses and programmes must implement the Two-Lane Approach to assessment by 2027, and its Cogniti AI agent platform is open for applications for Semester 2 / Q3 / Q4 2026. Otago’s AI Governance Policy, now in force, adds a university-wide governance layer that includes Te Tiriti obligations, data sovereignty, approval pathways, and formal committee oversight. (teachwell.auckland.ac.nz)
Overall State of Adoption
Schools and kura: use is mainstream, but capability and consistency still lag
The strongest national benchmark is still TALIS 2024. It shows that 63% of New Zealand Year 1–10 teachers had used AI in the previous year, and 69% of Year 7–10 teachers had used AI, compared with 36% across the OECD lower-secondary average. Among New Zealand teachers who used AI, the most common uses were generating lesson plans (78%) and learning about or summarising a topic (73%), while only 12% used AI to assess student work. TALIS also shows demand for capability-building remains high: 55% of secondary teachers and 36% of primary teachers had taken part in PLD on AI, yet 63% of secondary and 53% of primary teachers still reported needing PLD in AI for teaching and learning. (educationcounts.govt.nz)
NZCER’s recent primary-school AI study adds an important second layer. It found widespread experimentation by teachers, but also weak resourcing and uneven confidence: three-quarters of surveyed teachers had no school-funded premium AI access, fewer than half felt confident teaching responsible AI use, and 85% wanted more training. Student use was reported as higher outside school than inside school, and many students were unsure what the rules were. This reinforces the current picture: AI is present in classrooms, but institutional readiness is still catching up. (nzcer.org.nz)
The Ministry’s refreshed guidance points in the same direction. Its position is explicitly augmentation-first: AI should improve teaching and learning by supporting knowledgeable and skilled teachers, teachers must remain central, and AI-assisted marking must not replace professional judgment. That is a notable sign of maturity: the policy conversation has shifted from abstract warnings about ChatGPT to practical guidance on acceptable use cases, privacy, and assessment boundaries. (education.govt.nz)
AI literacy is moving from “initiative” to delivery infrastructure
The clearest current school-level literacy signal is Day of AI Aotearoa. According to Education Gazette, the localised pilot ran across nine schools and found a striking gap between exposure and understanding: 92% of students had tried AI tools, but almost none could explain how AI works before participating. After the programme, the share of students who felt able to explain AI rose from 20% to 64% in one primary setting and from 50% to 82% in one secondary setting; 93% recognised that AI can make mistakes or be unfair. Teachers also reported substantial confidence gains, and the programme explicitly integrated te ao Māori values and data sovereignty discussion. (gazette.education.govt.nz)
This matters because it suggests New Zealand’s most scalable school response is not currently a mandated AI subject, but a layer of AI literacy infrastructure built around free resources, teacher facilitation, and culturally localised framing. The same pattern appears in the AI Forum’s 2026 Blueprint, which describes Day of AI Aotearoa as launched in March 2026 and frames AI-enabled education as a national ambition rather than a fringe experiment. (gazette.education.govt.nz)
Tertiary education: from guidance to operating models
The tertiary sector is now the clearest example of institutional codification. At the University of Auckland, the Two-Lane Approach to assessment will be required across all courses and programmes by 2027. In this model, Lane 1 covers controlled assessments where authenticity and assurance of individual performance are critical, while Lane 2 allows uncontrolled conditions, including unrestricted AI use where appropriate. The University is explicit that the distinction is not simply “no AI” versus “full AI,” but a design choice tied to assessment purpose. From 2026 onwards, Lane 1 / Lane 2 terminology is expected to become visible to students. (teachwell.auckland.ac.nz)
The University of Auckland is also moving from policy to tooling. Its Cogniti platform is described as a generative AI chatbot tailored to course content and embedded inside Canvas, providing a “safe and controlled AI environment for coursework.” Applications are open for the second half of 2026, and the platform can be used for Socratic tutoring, role-play, formative feedback, and navigation of course materials. (teachwell.auckland.ac.nz)
Massey University remains aligned with the same general direction, using an Artificial Intelligence Use Framework that ranges from No AI through to AI planning, AI collaboration, and more expansive modes of permitted use. Its public guidance emphasises record-keeping, tool disclosure, and ethical use rather than blanket prohibition. (massey.ac.nz)
The University of Otago adds another sign of maturation: its AI Governance Policy, effective 10 March 2026, establishes a university-wide governance framework for AI across development, deployment, procurement, and use. The policy explicitly references Te Tiriti obligations, Māori data sovereignty, risk assessment before using AI on confidential or restricted data, and formal oversight by Council committees. That is less about classroom practice than about institutional readiness, but it is highly relevant: tertiary AI adoption is now being managed as a governance issue, not just a teaching issue. (otago.ac.nz)
System agencies: NZQA remains the most mature production user
The strongest example of operational AI in New Zealand education is still NZQA. NZQA says it used an AI-powered automated text scoring tool to mark more than 55,000 writing literacy assessments in May 2025, with experienced human markers double-checking over a third of results, especially near achievement boundaries. NZQA says this helped return results 3.5 weeks earlier than the previous year. (www2.nzqa.govt.nz)
NZQA has also continued the generative-AI upgrade of its chatbot Awhina, which began in April 2025. The system is designed to search public NZQA content, rank relevant results, and generate clearer answers with links to source material, with guardrails intended to protect privacy and prevent inappropriate learning from user inputs. (www2.nzqa.govt.nz)
A smaller but telling June update is that the NCEA site now has a dedicated Generative AI in NCEA assessment page, even though that page currently mainly routes users back to Ministry guidance. Its existence is still meaningful: AI has become established enough in assessment practice that it now warrants dedicated navigation inside the national qualifications support environment. (ncea.education.govt.nz)
Policy, Governance, and Curriculum
Ministry policy: clearer on marking, still principles-based overall
The Ministry’s current guidance remains principles-based in tone, but it is no longer vague. It says teachers must remain central, warns against entering personal information into AI tools, and now gives explicit marking guidance developed with NZQA. The guidance says AI use for marking is best suited to lower-risk assessment-for-learning contexts, that complex and summative uses require caution, and that NCEA internal assessment marking should not be delegated to AI. (education.govt.nz)
This is an important shift because it gives schools something closer to an operating boundary. New Zealand still does not have a single mandatory national school AI framework on the scale seen in some other jurisdictions, but it now has a clearer official position on where AI is appropriate, who remains accountable, and what kinds of educational decisions must stay human-led. (education.govt.nz)
Privacy and Māori data sovereignty are becoming even more central
The Office of the Privacy Commissioner continues to state that agencies should conduct a Privacy Impact Assessment before using generative AI tools, and specifically says agencies should engage with Māori about impacts on communities and taonga information. Its broader AI guidance also stresses the need to understand training-data quality, bias, and the privacy implications of using personal information throughout the AI lifecycle. (privacy.org.nz)
The wider public environment now makes those warnings more salient. In the Privacy Commissioner’s 2026 annual survey, 71% of New Zealanders said they were concerned about children’s privacy, and 67% were concerned about agencies or businesses using AI to make decisions about people using personal data. Māori respondents showed higher concern and lower trust across many measures. For education, that means trust and legitimacy are becoming harder constraints on AI rollout, not softer ones. (privacy.org.nz)
Curriculum direction: secondary AI timing is holding, junior timing has stretched
The biggest curriculum update since April is the 14 May 2026 rollout reset. Years 9–10 stay on track for full implementation from the start of 2027, but many Years 0–8 learning areas are delayed to 2029. In practice, that means AI-related curriculum development at the junior-secondary level remains live, while full system embedding across younger year levels will take longer. (tahurangi.education.govt.nz)
At the senior-secondary level, subject descriptors still show AI being embedded across subjects rather than isolated into a single compulsory AI subject. The published descriptors say Digital Technologies (Year 11) includes emerging technologies and their social, cultural, and ethical implications; Computer Science (Years 12–13) includes artificial intelligence and intelligent systems; and Media, Journalism & Communications addresses the impact of artificial intelligence on media and communication. That pattern still points to cross-subject integration. (newzealandcurriculum.tahurangi.education.govt.nz)
The Government’s 16 May 2026 qualification decisions reinforce the importance of this. The new Year 12 and 13 qualifications will be subject-based, and each subject will include coursework plus at least one exam. That broader redesign is not an AI policy in itself, but it creates a more formal environment in which questions of AI use, authenticity, and assessment design will need to be resolved subject by subject. (tahurangi.education.govt.nz)
Research Overview
Official system evidence
TALIS 2024 still provides the best national-scale evidence that New Zealand teachers are ahead of many international peers in AI use, but also far from settled in capability. New Zealand teachers are more likely than OECD peers to use AI for planning and content support, yet still report substantial need for professional learning and do not appear to be using AI heavily for high-stakes assessment judgment. (educationcounts.govt.nz)
NZCER: the capability gap remains one of the sector’s defining issues
NZCER’s primary-school AI study remains the most useful recent research addition to the New Zealand evidence base. Its value is not that it offers a nationally representative headline for all schools; it does not. Its value is that it shows what emerging practice looks like among teachers already relatively engaged with AI. That sample reveals both momentum and fragility: teachers are experimenting widely, but many lack premium tools, formal training, policy clarity, and confidence in teaching responsible use. (nzcer.org.nz)
The study also underscores a critical policy reality: student AI use is not primarily a school-contained phenomenon. Many students encounter AI first outside school, and often without adult guidance. That makes school policy increasingly about response and literacy, not simple control. (nzcer.org.nz)
ERO: monitoring is still active, not yet conclusive
ERO’s current work programme still includes the question “How are schools and students using AI?” but, as of 10 June 2026, it has not yet published a dedicated report answering it. That means practice is still running ahead of full system evaluation. (evidence.ero.govt.nz)
Case Studies
1) Day of AI Aotearoa / Westlake Girls’ High School
This is currently the strongest example of AI literacy at school-system edge. The programme was piloted in nine schools, localised for New Zealand, evaluated with NZCER involvement, and designed to work without specialist expertise. The reported gains in student explanation of AI, ethical awareness, and teacher confidence suggest that low-cost, teacher-led literacy models may be one of the fastest scalable responses available to schools. (gazette.education.govt.nz)
2) Aotea College and Hobsonville Point Secondary School
These remain important because they show how schools are redesigning assessment practice around AI rather than relying on detectors. The Ministry’s case studies highlight moves such as updated authenticity statements, verbal checks, version-history requirements, assessment checkpoints, and supervised components. They are still among the clearest examples of institution-level school governance in the New Zealand context. (education.govt.nz)
3) University of Auckland: Cogniti in coursework
The University of Auckland’s Cogniti rollout is one of the clearest tertiary examples of bounded, course-specific AI deployment. The platform is embedded in Canvas, steered by course materials, and positioned as a safer alternative to uncontrolled external chatbot use. One University case study describes a steered AI agent in BIOSCI 107 providing 24/7 Socratic support, immediate feedback, and a private space for conceptual inquiry. Another cross-faculty case study says the 2025 trial used Cogniti both as a course administrator and as a Socratic tutor, while giving teachers insight into common student questions. (teachwell.auckland.ac.nz)
4) NZQA: automated text scoring at national scale
NZQA’s writing-assessment scoring remains the most consequential live deployment because it sits inside a national assessment system, uses explicit human quality assurance, and has already been used at substantial scale. Compared with most school and university AI use, this is less exploratory and more operational. (www2.nzqa.govt.nz)
5) University-wide governance at Otago
Otago’s AI Governance Policy is a case study in institutional maturation rather than classroom innovation. It shows a university treating AI as a whole-of-organisation issue spanning teaching, research, procurement, risk, culture, privacy, and committee accountability. For the sector, that is a sign that governance is moving upward into core institutional architecture. (otago.ac.nz)
Emerging Trends
1) Augmentation still dominates substitution
Across schools, universities, and NZQA, the dominant pattern remains AI as support: planning, feedback, tutoring, search, service navigation, and bounded assistance. New Zealand’s official and institutional documents consistently preserve human accountability in teaching and assessment. (education.govt.nz)
2) Assessment redesign is becoming formal doctrine
What was emerging in early 2026 is now much clearer. In schools, the Ministry discourages AI-led marking for NCEA internal assessment. In tertiary education, the University of Auckland is building a formal controlled/uncontrolled assessment architecture, and Massey continues to specify permitted AI use at assessment level. The direction of travel is unmistakable: assessment design is replacing detector-led enforcement as the credible response to AI. (web-assets.education.govt.nz)
3) AI literacy is scaling faster than full curriculum integration
New Zealand’s fastest-moving AI response is not a compulsory standalone national AI subject. It is the growth of literacy programmes, localised resources, professional communities, and subject-embedded content. That includes Day of AI Aotearoa, Kōtui Ako offerings, Tāhūrangi AI resources, and AI-related content in senior subject descriptors. (gazette.education.govt.nz)
4) Unevenness is now the key implementation problem
The gap is no longer between “AI adopters” and “non-adopters” in any simple sense. It is between institutions with coherent governance, training, and tool choices, and those still relying on informal teacher improvisation. NZCER’s findings on training demand and limited tool access, combined with Day of AI’s emphasis on the AI literacy divide, make that unevenness one of the sector’s defining issues. (nzcer.org.nz)
5) New Zealand’s distinctive feature remains cultural legitimacy
Te Tiriti obligations, Māori data sovereignty, Pacific values, and cultural responsiveness are not peripheral in the New Zealand AI-in-education discussion; they are increasingly core to it. That is visible in Ministry and Privacy Commissioner guidance, in Day of AI localisation, in the Otago governance framework, and in wider ecosystem documents such as the AI Forum Blueprint. (privacy.org.nz)
Conclusion
As of 10 June 2026, AI in education in New Zealand is best understood as normalised in use, still uneven in capability, and significantly more mature in governance than it was even a few months ago. The strongest changes since early April are not flashy product launches. They are institutional: clearer Ministry guidance, harder-edged assessment models, more explicit qualification and curriculum timelines, and more visible governance frameworks at major universities. (education.govt.nz)
The sector’s centre of gravity is also clearer now. New Zealand is not moving toward an “AI replaces educators” model. It is moving toward a model in which AI is allowed to support planning, literacy, tutoring, service delivery, and selected assessment operations, while human judgment, transparency, privacy, and cultural legitimacy become stricter requirements. That is evident in school guidance, university assessment design, and NZQA’s supervised deployment model. (web-assets.education.govt.nz)
The defining challenge from here is implementation quality. The evidence suggests New Zealand has crossed the threshold from experimentation to operational adoption, but it still has not reached a fully coherent national operating model. The institutions furthest ahead are those treating AI as a whole-system education design issue involving pedagogy, assessment, governance, data, equity, and trust all at once. (educationcounts.govt.nz)