Teachers use AI responsibly for marking and feedback by keeping it a drafting tool, not the decision-maker: AI can suggest feedback against your rubric and success criteria, but you verify every judgement, protect student privacy, and stay accountable for grades under NCEA and NZQA moderation.
What does responsible AI use in marking actually mean?
Responsible AI use in marking means keeping AI a tool that drafts, while you stay the professional who decides. A model can suggest comments, point to where a response meets or misses a success criterion, and take the grind out of repetitive marking. But the grade, the reasoning behind it, and the accountability for it remain yours. AI informs your judgement; it does not replace it.
In practice this rests on three commitments. First, verify: read the work yourself and check every AI-generated judgement against what the student actually wrote. Second, be transparent with students, whanau and colleagues about how and where AI was used. Third, keep a human in the loop for any decision that carries weight, so no automated output goes unchecked before it reaches a report or a moderation sample.
The distinction matters because marking is not just scoring. It is a professional act that draws on your knowledge of the learner, the class and the curriculum. A tool that lightens the mechanical load frees you for that judgement work rather than substituting for it.
Which marking tasks suit AI, and which should stay with you?
AI suits low-stakes, formative marking best: draft feedback on practice essays, first-round comments on a paragraph, or a quick consistency check against your rubric. High-stakes decisions, especially summative NCEA judgements, should stay firmly with you, because you carry the accountability and the contextual knowledge of each learner that a model lacks.
A useful test is to ask what happens if the AI is wrong. On a formative draft, a wrong suggestion costs a few minutes of your review. On a final internally assessed grade, it could misreport a student’s achievement. Match the level of human oversight to the stakes: the higher the consequence, the more of the marking you should do and verify yourself.
A practical way to build confidence is to pilot AI on a set of scripts you have already marked. Compare its draft comments with your own, note where it agrees and where it drifts, and only widen its use to live drafts once you trust its behaviour on your subject and year level. That trial period also shows students and colleagues that you adopted the tool deliberately rather than on faith.
- Well suited: formative feedback on drafts, comment-consistency checks, and surfacing where a response misses a criterion.
- Use with caution: a first-pass mark against a rubric that you then review in full before it counts.
- Keep human-led: final NCEA achievement levels, moderation samples, and any high-stakes report grade.
How do you protect student data and privacy when using AI?
Protect student privacy by never pasting identifiable student information into a general-purpose AI tool, and by choosing tools that are clear about how they store and use data. Student work is personal information, and under New Zealand’s Privacy Act 2020 schools are responsible for how it is collected, shared and stored. Treat any AI service as a third party you are entrusting with that data.
Before you upload anything, check who can see it, whether it is used to train models, and where it is held. Where possible, remove names and identifying details, or use a tool designed for education that keeps data under your control rather than a consumer chatbot repurposed for marking.
If you are unsure whether a tool meets your obligations, ask your school’s privacy officer or leadership team before you use it, not after. Many schools now keep a short list of approved services precisely so individual teachers are not left assessing data-handling risk alone. Recording which tool you used, and for what, also makes any later privacy query straightforward to answer.
- Avoid pasting student names or identifying details into consumer AI chatbots.
- Prefer tools that state clearly whether your data is used to train their models.
- Check your school's privacy policy and Privacy Act 2020 obligations before uploading work.
- Keep the teacher, not the tool, as the owner of the final record.
Can AI make feedback more consistent without replacing your judgement?
Yes. AI can make feedback more consistent by applying the same rubric, success criteria and comment bank to every student, which is exactly where fatigue and time pressure usually cause human marking to drift. When the model drafts against your criteria, the twentieth script gets the same careful attention as the first, and you edit from a consistent baseline rather than a blank page.
The judgement still belongs to you. The value is in the first draft: a structured, criterion-referenced comment you can accept, sharpen or overrule in seconds. You add the nuance a model cannot reach, such as the student’s prior learning, the specific misconception behind an error, or the encouragement that lands for that particular learner.
Consistency, used this way, is not about making every comment identical. It is about making sure your standard is applied evenly, so feedback is fair across a class and defensible if a grade is later questioned.
What do NCEA and NZQA rules mean for AI-assisted marking?
For NCEA, AI can support your marking but it cannot satisfy NZQA’s requirements on your behalf: you remain responsible for authenticity, for the assessor judgement, and for internal moderation. NZQA sets clear expectations that internally assessed work is genuinely the student’s own and that grading decisions are defensible against the achievement standard.
That has two implications. AI-assisted feedback on drafts is fine, but you must still confirm the final submission is the student’s authentic work. And when your grades are moderated, internally or by NZQA, the reasoning has to hold up as yours, not an unexamined model output. Keeping AI a drafting aid, with your verification on top, is what keeps you aligned with those rules.
It helps to keep a light record of how AI featured in your process: which drafts it commented on, and how you checked the final judgement. You do not need a lengthy audit trail, but a few notes make moderation smoother and let you show, if asked, that the achievement decision was reasoned and made by you.
How can a tool like JeddAI help you mark responsibly?
A purpose-built tool like JeddAI helps you mark responsibly by drafting feedback against your own rubric, success criteria and comment banks, then leaving every decision to you. You connect or upload student work, JeddAI produces criterion-aligned feedback and marking, and you review, edit or overrule it before anything is final, so you save time while staying in control.
That design matches the principles above: the teacher stays the decision-maker, feedback is anchored to your criteria rather than a generic model opinion, and consistency comes from your standards rather than the tool’s guesswork. If you want to try this approach, you can Get started with JeddAI and keep your professional judgement at the centre of every grade.
| Aspect | Responsible use | Risky use |
|---|---|---|
| Who decides the grade | The teacher, after reviewing the draft | The AI, output accepted unchecked |
| Student data | Identifiers removed or kept in a controlled tool | Named work pasted into a consumer chatbot |
| Alignment | Anchored to your rubric and success criteria | A generic model opinion with no criteria |
| High-stakes NCEA judgements | Human-led and verified for authenticity | Delegated to the model to save time |
| Transparency | Students and colleagues know AI assisted | Use hidden from students and moderators |
Frequently asked questions
Is it cheating for a teacher to use AI to help mark?
No, provided you use it as a tool and stay accountable. If AI drafts feedback, you verify it against the work, and you make the final decision, it is a legitimate aid rather than a shortcut around your judgement.
Can AI decide a final NCEA grade on its own?
No. You remain the assessor responsible for the judgement and for authenticity, and NZQA moderation expects the reasoning to be defensibly yours. AI can draft, but the achievement decision must be human-made.
Should I tell students I use AI to help with feedback?
Yes. Being transparent about how and where AI assists your marking builds trust, models responsible AI use for students, and keeps you consistent with your school's expectations.
How do I check AI-generated feedback is accurate?
Read the student's work yourself and compare each AI comment against the rubric and the actual response. Treat the draft as a starting point to confirm or correct, never as a finished mark.
Is it safe to upload student work to AI tools?
Only if the tool handles data responsibly. Prefer education-focused tools that are clear about storage and training use, follow your school's privacy policy, and remove identifying details where you can.
Get started with Jeddle
Jeddle gives teachers and students instant, syllabus-aligned feedback powered by JeddAI.
Looking for study material? Browse Jeddle's Australian-English subject resources, or explore more articles on AI in Education.
