Transparency
How WriteWise uses AI
What the AI in WriteWise does, how to use it well and safely, how we protect your work, and every change we make to it.
How the AI works
WriteWise uses AI models from OpenAI, through OpenAI’s business service. Everywhere WriteWise uses AI, the screen says so. We use it for a few specific jobs:
- Feedback for students. Feedback and annotations on essays and reports, linked to the task, the NCEA standard and its criteria. It points to what to work on next. It does not rewrite the work or give you text to hand in.
- Marking support for teachers. A suggested grade or mark with reasons, based on the standard, its criteria and exemplars. It is a suggestion only. The teacher decides every grade.
- Reading handwriting. Photos of handwritten work are turned into text that you check before it is used.
- Teacher tools. Drafts of assignments, rubrics, attachment summaries and digital exams made from PDFs, for the teacher to review and edit before students see them.
The AI is sent the work, the task and the curriculum context. Your school and teacher decide whether AI is used. A student chooses when to ask for feedback.
Getting good results
- Write your own draft first. Feedback is most useful on real work.
- Read the feedback against the task and the criteria, and decide what to change yourself.
- Ask again after you revise, to see whether your changes addressed the feedback.
- Teachers: treat AI marking as a second opinion. Check it against the work before you rely on it.
For students: using AI safely
- Focus on the task your teacher set.
- Don’t put personal information in your work, such as your name, address or phone number.
- The AI can be wrong. If something doesn’t look right, ask your teacher.
- The work you hand in must be your own. Use the feedback to improve it.
- If you are going through a hard time and write about it, WriteWise may show you where to get support instead of feedback. You can always talk to your teacher or another adult you trust.
Mistakes and bias
AI can make mistakes. Its output can be inaccurate or incomplete, and it can show bias. To reduce this, feedback and marking are grounded in the official standard, its criteria and exemplars, and names and other identity details are left out of what the AI sees. A teacher makes every assessment decision that reaches a student, and teachers choose which AI comments and feedback students see, sharing only what they believe is helpful.
Your privacy
- Work you submit is not used to train AI models. OpenAI is not permitted to use it for training.
- OpenAI processes the work in the United States. Our Privacy Policy explains what we collect, where it is stored, and the safeguards for overseas processing.
Your responsibilities
- Students: follow your school’s rules and NCEA’s conditions of assessment. Work you submit for assessment must be your own.
- Teachers and schools: review AI suggestions before using them, make the final assessment decision, and tell students and parents that your school uses WriteWise and its AI features.
- Everyone: use WriteWise in line with our Terms of Use and the Privacy Act 2020.
Reporting a problem
If AI output is wrong, biased or inappropriate, tell us through the feedback button in WriteWise or email leon@writewise.co.nz. Students can also tell their teacher. We investigate every report, and where needed we pause the feature and tell the affected schools.
Current models
| Features | Model | Provider |
|---|---|---|
| Feedback and marking Feedback on essays and reports, marking, annotations, grade estimates, finding quotes, reading handwriting, and generating assignments. | GPT-6.1 | OpenAI |
| Importing and reading documents Importing exam papers and rubrics from PDF, and reading scanned exam booklets. | GPT-6 Luna | OpenAI |
Change history
When we change the model behind a feature, we list it here and show a notice beside that feature in WriteWise for up to 30 days.
3 Nov 2026: Importing and reading documents
Introduced GPT-6 Luna. Importing exam papers and rubrics from PDF, and reading scanned exam booklets, use GPT-6 Luna.
18 Oct 2026: Feedback and marking
Moved from GPT-5.4 (most features), with GPT-5.2 and GPT-4o for some to GPT-6.1. All AI feedback and marking moved to GPT-6.1.
