Choose an AI grading tool that marks against your own achievement standards, marking schedules and criteria rather than a generic scale. Prioritise tools that accept your standard, keep judgements standard-referenced, produce specific feedback, protect student data, and leave the final grade decision with you as the teacher.
What makes NCEA marking different from other grading?
NCEA marking is standard-referenced, so a grading tool has to judge each piece of work against a published achievement standard rather than rank students or award a percentage. Each standard sets out what Achieved, Merit and Excellence look like, and a marking schedule shows how evidence maps to those levels. A tool that ignores this and applies a generic scale will produce grades that do not hold up.
This matters most for internal assessment, where your judgements are moderated for consistency against the standard. Any AI you bring in should reinforce that logic, not work around it. The right question is not can it grade but can it grade against this specific standard and its criteria, and can it explain its judgement in terms a moderator would recognise.
It also shapes what good feedback looks like. In a standard-referenced system, a comment is most useful when it names the descriptor a student has met or is reaching for and points to the specific evidence in their work. A tool that grasps this produces feedback a student can act on and a colleague can verify, which is exactly what you want when a grade might be sampled during moderation.
Which capabilities actually matter for standard-based marking?
The capabilities that matter most are the ones that tie a draft grade to the standard: accepting your own achievement standard, marking schedule and exemplars, and expressing judgements as Not Achieved, Achieved, Merit or Excellence rather than a number. A tool that only outputs a score forces you to translate it back into standard language yourself, which removes most of the time saving.
After alignment, look at feedback quality and control. Strong tools generate specific comments linked to the descriptor a student did or did not meet, and let you edit every judgement before it reaches a learner. Weaker tools produce generic praise or vague suggestions that could apply to any subject.
- Accepts your own standard, marking schedule and criteria as the marking reference.
- Expresses grades in NCEA terms (Not Achieved, Achieved, Merit, Excellence) with reasoning.
- Produces specific feedback tied to descriptors, not generic comments.
- Lets the teacher review and change every judgement before it is finalised.
- Keeps student data secure and clear about where work is stored.
Should AI make the final grading decision?
No. The teacher should always make the final grading decision, and a good AI tool is built around that principle. For internal NCEA assessment your professional judgement is what is moderated, so the workflow should have AI draft a grade and feedback that you then confirm, adjust or override. A tool that finalises grades automatically, or makes it hard to change its judgement, is a poor fit for a standard-referenced system.
This is not only a compliance point, it is a quality one. AI can miss nuance in a borderline Merit or Excellence response, or over-credit surface features. Keeping yourself in the loop means the marking schedule, not the model, decides the boundary. Treat AI output as a well-organised first pass that surfaces evidence for you to check.
How do you check data privacy before adopting a tool?
Before adopting any AI marking tool, confirm where student work is stored, who can access it, and whether the tool is transparent about how submissions are handled. New Zealand schools have obligations around student data, so treat privacy as a filter you apply before you ever look at features. If a vendor cannot answer these questions plainly, that is a signal in itself.
It is also worth checking your school or kura policy on AI use and student consent, and whether the tool lets you avoid uploading identifying details. A tool that works from the writing itself, without needing extra personal data, is easier to adopt responsibly and easier to explain to whanau if asked.
A practical test is to read the vendor’s privacy information the way a cautious parent would. Can you tell, in plain language, where the work goes, how long it is kept, and whether it is used for anything beyond marking your students’ work? If those answers are clear and reasonable, the tool clears a bar that many generic AI products do not, and you can move on to judging its marking on merit.
What questions should you ask a vendor before committing?
Ask whether the tool can mark against your own achievement standard and marking schedule, how it expresses a judgement, and how much control you keep over the final grade. These three answers separate tools designed for standard-referenced marking from generic essay scorers that were built for a different system and rebadged.
Then pressure-test it on real work. Run a set of scripts you have already marked and compare the tool’s drafts against your judgements, paying attention to borderline cases and the quality of the feedback. A short trial on your own standards tells you more than any feature list, because it shows whether the tool understands the descriptors or just recognises good writing in general.
How does a tool like JeddAI approach NCEA-aligned marking?
JeddAI is built so that marking runs against your own standards rather than a fixed scale. You connect or upload student work, and JeddAI drafts feedback and marking aligned to the rubric, success criteria and comment banks you provide, so the output reflects the descriptors you mark to instead of a generic model of quality.
Crucially, the teacher stays in control. JeddAI produces a first pass that you review and edit before anything reaches a student, which fits how internal assessment and moderation work in New Zealand and saves marking time without handing over the judgement. If you want to see how that feels on your own standards, you can Get started with JeddAI and trial it on a set of scripts you have already marked.
| What to look for | Why it matters for NCEA | Warning sign |
|---|---|---|
| Marks against your own standard | NCEA is standard-referenced, so the tool must use the achievement standard and marking schedule | Only offers a fixed rubric or percentage score |
| Uses NCEA-level judgements | Grades should read as Not Achieved, Achieved, Merit or Excellence with reasoning | Outputs a raw number you have to translate yourself |
| Specific, criterion-linked feedback | Feedback tied to descriptors supports student progress and moderation | Generic praise that could apply to any subject |
| Teacher keeps the final decision | Your professional judgement is what gets moderated | Auto-finalises grades or makes overrides hard |
| Clear data handling | Schools carry obligations around student information | Vague about where student work is stored or who sees it |
Frequently asked questions
Can an AI tool mark internal NCEA assessment for me?
It can draft a grade and feedback, but you should make the final judgement. Internal assessment is moderated against your professional decision, so use AI as a first pass you confirm or adjust, not as the final marker.
Does the tool need to know it is marking NCEA?
What matters more is that it marks against your achievement standard, marking schedule and exemplars. A tool that accepts those and expresses judgements in NCEA levels will fit far better than one applying a generic scale.
How do I know if an AI grade is trustworthy?
Trial it on scripts you have already marked and compare, paying close attention to borderline Merit and Excellence responses. Consistency on your own standards is a better test than any vendor claim.
Is it safe to upload student work to an AI grading tool?
Only after you check where the work is stored, who can access it, and your school or kura policy on AI and student data. Favour tools that are transparent and do not require extra identifying details.
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 EdTech & Tools.
