AI in Education in New Zealand: A Living Whitepaper

Executive Snapshot

AI adoption in New Zealand education is moving from individual experimentation to structured, curriculum-linked implementation.

The most significant developments since 13 July 2026 are:

• The Government has approved Applied Intelligent Systems, a new senior-secondary subject covering artificial intelligence, machine learning, agents, large language models, workflow automation, and human oversight.
• Ministry-funded Year 9–10 mathematics resources now include access to an AI-powered learning tool through Education Perfect, alongside curriculum-aligned lessons, assessment tools, learner analytics, and inclusive accessibility features.
• AI professional learning is becoming a more formal education market, with the University of Canterbury offering a 20-hour short course on authentic assessment in the age of AI from 17 August 2026.
• The national AI-in-education ecosystem is becoming more coordinated through the AI Forum’s refreshed Blueprint, a growing educator community of practice, new assessment workstreams, and the 2026 New Zealand AI in Higher Education Symposium.
• The overall adoption pattern remains uneven: teachers are using AI widely for planning and content development, but school-level capability, policy clarity, privacy safeguards, and access to approved tools remain inconsistent.

The sector is not converging on an “AI replaces teachers” model. The emerging New Zealand approach is human-led, assessment-aware, culturally grounded, and increasingly based on bounded tools rather than unrestricted chatbot use.

What Has Changed Since 13 July 2026

1. AI has entered the future senior-secondary subject pathway

On 6 August 2026, the Ministry of Education and Tāhūrangi announced nine industry-led subjects for the new senior-secondary curriculum. One of these is Applied Intelligent Systems. (tahurangi.education.govt.nz)

The proposed subject will teach students to:

• Design, create, use, and evaluate intelligent systems.
• Work with artificial intelligence and machine learning.
• Use no-code and low-code platforms.
• Apply workflow automation, agents, and large language models.
• Identify where AI systems may fail.
• Understand bias, limitations, and the continuing need for human oversight.
• Solve real-world problems through project-based individual and group work.

The subject is intended to require no prior programming experience and to operate on standard school computers. Development will occur during 2026 and 2027, assessment design is scheduled for 2027, piloting for 2028, and access for Year 12 and Year 13 students is planned from 2029 and 2030 respectively through the new NZCE and NZACE qualifications. (tahurangi.education.govt.nz)

Why it matters: AI is moving from an optional topic within Digital Technologies into a visible senior-secondary pathway connected to workforce preparation.

2. AI is becoming embedded in Ministry-funded digital mathematics provision

The Ministry’s Year 9–10 digital mathematics platform, delivered through Education Perfect, is available from Term 1 2026 and includes an AI-powered tool alongside structured lessons, assessment tools, progress insights, text-to-speech, alt-text, and multimodal presentation. The platform is optional for schools, with professional learning provided by Education Perfect. (newzealandcurriculum.tahurangi.education.govt.nz)

The platform’s AI functions include:

• AI-supported feedback on extended responses.
• Reading assistance and vocabulary support.
• Hints and guided learning loops.
• AI-supported worksheet and activity creation.
• Learner-progress reporting for teachers.

This represents an important shift from teachers independently accessing general-purpose AI tools to schools receiving AI functionality through a curriculum-aligned platform with an education-provider relationship.

3. Professional learning is becoming more specialised

The University of Canterbury’s UC Online is now offering Designing Authentic Assessment in the Age of Artificial Intelligence, a 20-hour course aimed at secondary teachers, tertiary educators, and education leaders. The course focuses on AI-supported grading and feedback, the limitations of automated assessment, responsible-use guidelines, and authentic assessment design. (uconline.ac.nz)

The development is significant because it treats AI capability as a professional discipline rather than a short-term technology orientation. It also reflects the sector’s growing emphasis on:

• Assessment redesign.
• Teacher judgement.
• Feedback quality.
• Academic integrity.
• Practical implementation rather than generic AI awareness.

4. National coordination is becoming more visible

The AI Forum’s refreshed 2026 Blueprint for Aotearoa places education within a national strategy focused on capability, adoption, trust, social licence, data sovereignty, and sustainable AI. (aiforum.org.nz)

Its AI in Education workstream now includes groups focused on:

• Public-private partnerships and teacher professional learning.
• Culturally aligned AI.
• AI and assessment.
• Defining AI literacy.
• School-leader guidance.
• AI professional development.

The associated national AI-in-education community of practice has grown to more than 580 members, according to the AI Blueprint. (aiforum.org.nz)

The University of Otago is also hosting a 2026 New Zealand AI in Higher Education Symposium featuring practitioner exchange, an Agentic Innovation Awards programme, and an interactive immersion lab. The event signals that tertiary institutions are beginning to treat AI adoption as a sector-wide professional and institutional issue. (otago.ac.nz)

Current State of Adoption

Schools and kura: widespread experimentation, uneven capability

The strongest national benchmark remains TALIS 2024. It found that:

• 69% of New Zealand Year 7–10 teachers had used AI in the previous year, compared with 36% across the OECD lower-secondary average.
• Among New Zealand teachers who had used AI, 78% had used it to generate lesson plans or activities.
• 73% had used it to learn about or summarise a topic.
• Only 12% had used AI to assess or mark student work.
• Secondary teachers reported more AI-related professional-learning participation than primary teachers, but also a higher level of unmet need. (educationcounts.govt.nz)

The data suggests that AI adoption is already mainstream at the level of teacher workflow, but not yet at the level of high-stakes assessment or systematic learning analytics.

NZCER’s focused primary-school study provides additional detail. Among its cohort of 266 teachers and 147 students:

• Teachers commonly used AI for lesson planning, assessment design, and personalising learning materials.
• Three-quarters of responding teachers had no school-funded premium AI access.
• Fewer than half felt confident teaching responsible AI use.
• 85% wanted more training.
• More than half of surveyed students reported using generative AI, with use more common outside school than inside school.
• Many students were uncertain about their school’s rules. (nzcer.org.nz)

Because the NZCER teacher sample was disproportionately interested in AI, it should not be treated as a national prevalence estimate. Its value is in showing the practical conditions surrounding adoption: experimentation is occurring faster than policy, training, and resourcing.

Secondary education: moving from general AI use to curriculum-linked capability

Secondary education is where the most visible institutional shift is occurring.

The proposed Applied Intelligent Systems subject gives schools a future pathway for teaching AI as a practical, critical, and workforce-relevant capability. It is explicitly designed to combine technical experimentation with evaluation, accountability, and human judgement. (tahurangi.education.govt.nz)

At the same time, AI is being introduced through digital subject platforms such as the Ministry-funded mathematics provision. This creates a more bounded model in which AI is connected to:

• Specific curriculum outcomes.
• Teacher-controlled activities.
• Learner progress data.
• Structured feedback.
• Accessibility and differentiated learning.

This is materially different from informal use of public chatbots, although schools will still need to assess vendor privacy, data retention, model performance, cultural fit, and procurement arrangements.

Tertiary education: governance and assessment redesign are maturing

The University of Auckland’s two-lane assessment model remains one of the clearest examples of institutional adaptation:

• Lane 1 uses controlled conditions where AI is restricted by default.
• Lane 2 permits AI use in uncontrolled assessments, with students remaining responsible for the work they submit.
• From 2027, all courses are expected to identify assessments as Lane 1 or Lane 2. (auckland.ac.nz)

The model is important because it moves away from trying to prohibit AI in every learning context. Instead, it separates the need to authenticate core knowledge from the need to prepare students to use AI critically and responsibly.

The University of Otago’s AI Governance Policy, effective from 10 March 2026, applies across procurement, development, deployment, research, teaching, and operations. It embeds privacy, bias, security, cultural appropriateness, Te Tiriti o Waitangi, Māori and Pacific data sovereignty, staff capability, and human oversight into the institution’s governance framework. (otago.ac.nz)

University-level adoption is therefore increasingly characterised by:

• Institution-wide policy.
• Formal risk classification.
• Course-level assessment rules.
• Student AI literacy.
• Approved tools and learning environments.
• Greater attention to cultural and data governance.

National assessment: NZQA remains the most mature production deployment

NZQA’s automated text scoring for literacy writing assessments remains the strongest example of AI operating at national education scale.

In 2025, more than 55,000 writing assessments were marked using automated text scoring, with results returned approximately 3.5 weeks earlier than the previous year. Responses near the achievement boundary were check-marked by experienced human markers. NZQA reported that its earlier pilot of more than 36,000 writing samples produced results comparable to human marking. (www2.nzqa.govt.nz)

Budget 2026 provided a further $2.1 million for NZQA to pilot artificial intelligence and machine learning in national qualifications, including marking, moderation, and exam development. (web-assets.education.govt.nz)

The policy distinction remains important: NZQA’s controlled deployment is not evidence that AI should independently replace teacher or assessor judgement across the wider education system. It is evidence that AI can be used in high-volume assessment when supported by:

• Carefully bounded use cases.
• Domain-specific training data.
• Human check-marking.
• Quality assurance.
• Clear accountability.
• An ability to examine boundary cases.

Research and Evidence Base

AI use is ahead of AI capability

TALIS shows high teacher exposure but continued professional-learning demand. Most teachers were using AI for preparation and information work, while relatively few were using it for marking or learner-performance analysis. (educationcounts.govt.nz)

The central evidence-based conclusion is therefore:

New Zealand education has moved further in using AI tools than in building consistent systems for evaluating, governing, and teaching about AI.

AI literacy is being reframed as critical judgement

The University of Canterbury’s Scaffolded AI Literacy framework, or SAIL, defines AI literacy as more than knowing how to operate tools. It combines:

• Understanding core AI concepts.
• Practical tool capability.
• Critical evaluation.
• AI citizenship.
• Contextual and age-appropriate progression.
• Māori and Pacific perspectives.

SAIL is designed for both educators and learners and does not assume that AI should always be used. Its central question is whether a tool should be used in a particular context, not simply whether it can be used. (canterbury.ac.nz)

This approach aligns closely with the Applied Intelligent Systems subject and the AI Forum’s education workstream.

National evaluation is still catching up

The Education Review Office has identified “How are schools and students using AI?” as an active research question. The project is intended to examine current use by teachers, school leaders, and students, as well as benefits, challenges, and emerging impacts. (evidence.ero.govt.nz)

There is currently no new nationally representative 2026 adoption survey that supersedes TALIS. TALIS will not run internationally again until 2030, making the ERO work important for filling the near-term evidence gap. (educationcounts.govt.nz)

Governance, Privacy, and Cultural Legitimacy

The Ministry’s guidance continues to establish a clear hierarchy:

• Teachers remain responsible for student learning.
• Human oversight is required.
• AI should support, not replace, teacher professional judgement.
• Schools must have policies covering acceptable GenAI use.
• Student work submitted for assessment must be the student’s own.
• GenAI use is not permitted in NCEA external assessments. (education.govt.nz)

The Ministry also warns that AI systems may be weak in mātauranga Māori, te reo Māori, Pacific languages, and Polynesian cultures. It advises schools not to enter personal, confidential, sensitive, or copyright-protected information into public AI tools. (education.govt.nz)

The Privacy Commissioner’s education-sector guidance reinforces the operational implications. Education providers remain responsible for information collected through third-party platforms and should consider:

• Whether vendors use learner information to train AI systems.
• How long data is retained.
• Whether information is used for secondary commercial purposes.
• Whether a privacy impact assessment is required.
• How learners and parents are informed. (privacy.org.nz)

New Zealand’s distinctive governance challenge is therefore not only accuracy or cybersecurity. It is also cultural appropriateness, Māori data sovereignty, children’s rights, and community trust.

Case Studies

Case Study 1: Applied Intelligent Systems

A new senior-secondary subject will give students hands-on experience designing and evaluating AI-enabled systems. Its emphasis on no-code and low-code tools makes participation possible without requiring every student to become a programmer, while its focus on failure, accountability, and human oversight guards against purely instrumental “prompting” education. (tahurangi.education.govt.nz)

Case Study 2: Ministry-funded digital mathematics platform

The Education Perfect platform demonstrates how AI is being introduced through a bounded, curriculum-linked environment. Its intended uses include feedback, reading support, hints, learner progress insights, and teacher-created activities. The platform is optional and accompanied by professional learning, but it creates a model for scaling AI through centrally supported digital infrastructure. (newzealandcurriculum.tahurangi.education.govt.nz)

Case Study 3: NZQA automated text scoring

NZQA shows how AI can be deployed in high-volume assessment while retaining human oversight for boundary cases and accountability. The case is notable less because of automation alone than because it combines automation with staged testing, human moderation, and quality assurance. (www2.nzqa.govt.nz)

Case Study 4: University of Auckland assessment redesign

The two-lane model provides a practical institutional response to generative AI. It protects controlled demonstrations of core learning while allowing students to practise responsible AI use in other assessment settings. (auckland.ac.nz)

Case Study 5: University of Otago governance

Otago’s policy demonstrates the transition from informal guidance to institution-wide AI governance. It integrates academic, operational, legal, cultural, privacy, environmental, and capability considerations into a single framework. (otago.ac.nz)

Emerging Trends and Outlook

1. Bounded AI will outperform unrestricted chatbot adoption

The strongest implementation models now connect AI to a course, subject, assessment task, or institutional workflow. This improves accountability and makes it easier to define acceptable use, data boundaries, and human responsibilities.

2. Assessment redesign will remain the central strategic response

New Zealand is increasingly moving away from detector-led enforcement. The more durable approach is to combine:

• Controlled assessments.
• Authentic tasks.
• Oral or practical demonstrations.
• Process checkpoints.
• Disclosure requirements.
• Human observation.
• AI-permitted activities where appropriate.

3. AI literacy will be embedded across the curriculum

The likely direction is not a single compulsory AI subject for every learner. Instead, AI literacy will appear through Digital Technologies, English, mathematics, social sciences, vocational subjects, assessment practice, and broader citizenship education. Applied Intelligent Systems is likely to become the most explicit specialist pathway, while SAIL represents the broader cross-curricular approach.

4. The capability divide will become more important than the adoption divide

Most educators are no longer deciding whether AI exists. They are deciding whether they have:

• Suitable tools.
• Clear policies.
• Time for professional learning.
• Access to reliable data.
• Confidence in evaluating outputs.
• Support for culturally responsive implementation.

Schools and institutions with stronger governance and digital infrastructure are likely to gain more value from AI than those relying on individual teacher initiative.

5. Human judgement and cultural legitimacy will remain non-negotiable

The New Zealand operating model is likely to remain augmentation-first. The most acceptable uses of AI will be those that improve feedback, accessibility, planning, differentiation, and administrative efficiency without obscuring responsibility or weakening relationships between teachers and learners.

Conclusion

As of 18 August 2026, AI in New Zealand education is best characterised as widely used, increasingly structured, and still unevenly governed in practice.

The most important new development is the beginning of a transition from AI as an informal classroom tool to AI as part of the education system’s formal infrastructure:

• A senior-secondary Applied Intelligent Systems pathway is being developed.
• Ministry-funded mathematics resources include AI functionality.
• Universities are redesigning assessment and formalising governance.
• NZQA is expanding carefully controlled national assessment pilots.
• AI literacy and professional learning are developing into recognised fields of work.

The sector’s main challenge is no longer awareness. It is implementation quality.

New Zealand has a credible emerging model: AI can support teachers, provide timely feedback, extend accessibility, strengthen learning pathways, and assist selected assessment processes. However, scaling those benefits will depend on sustained investment in teacher capability, equitable access, privacy protection, Māori and Pacific data governance, culturally responsive design, and assessment systems that preserve authentic evidence of learning.

The direction of travel is clear: AI will become more present in New Zealand education, but the institutions most likely to succeed will be those that treat it as a whole-system design and trust challenge—not simply as a new software category.