Writing and Enforcing a School AI Academic Integrity Policy

Last updated: 2026-07-24

As three subjects and six teaching demonstrations bring AI into the classroom, another question walks in with them: students are using AI on their homework too. Without written rules, teachers each go their own way, and the same AI-assisted assignment loses marks in one class and passes in another. A clear academic integrity policy exists to fill that gap.

Why write one now

Detection tools are unreliable, and that needs saying up front. The AI content detectors on the market get it wrong often. They flag work students wrote themselves as AI, and they miss text that really was generated. Build a policy on "catch them and punish them" and you have built it on sand.

A safer starting point is that the rules are open. Students know how far they can use AI on each kind of task, and teachers know how to mark it and how to handle a case they suspect. The point of the policy is not to frighten anyone. It is to give everyone the same measuring stick.

Sort allowed from not allowed by assessment type

There is no single school-wide answer to "can I use AI", because it depends on what the assessment is trying to measure. The same tool is an assistant while you gather material and a ghostwriter in an exam that tests writing. A practical policy splits use into a few tiers and matches each kind of assessment to one of them.

Fully allowed: low-risk tasks meant for learning, such as using AI to explain a concept, generate practice questions, or translate vocabulary. These uses should be encouraged, and they can feed into student AI literacy activities.
Allowed with conditions: AI may help with brainstorming or checking, but the finished work must be the student's own, and they must declare which tools they used and what for, as when organising material for a project.
Not allowed: assessments that measure a student's own ability directly, such as graded essays, school exams, and assessments that count towards banding. Having AI write these is the same as plagiarism.

Shift the weight of assessment onto the process

When the finished piece is hard to tell apart from AI output, the assessment has to make the process visible. Asking students to hand in drafts and revision notes, to write a first draft under time in class, or to give a short oral defence of their work, all move the focus from "how good is this essay" to "how much does this student understand". The oral defence works especially well. Three questions usually settle whether someone knows the material.

This shift helps the teaching itself. It forces assessment design to answer a basic question: what am I actually trying to measure? Once that answer is clear, whether AI is involved stops being a threat and becomes one known factor you design around.

Student declarations and staff consistency

A policy needs two pillars to stand. The first is the student declaration. On set assignments, ask students to note briefly whether they used AI and for which part. The value of the declaration is not detection but teaching. It makes students pause each time they hand work in and think about where the line sits.

The second pillar is staff consistency. If teachers in different subjects read the same rule differently, the policy is worth nothing. We suggest the AI implementation team takes the policy to each subject panel for discussion, reaches agreement on the common situations, and writes the steps for handling a suspected case into a simple flow, so every teacher has the same approach in hand.

A policy skeleton you can adapt

Scope: which year groups, which subjects, and which kinds of assignment and assessment the policy covers.
Tiers of use: fully allowed, allowed with conditions, and not allowed, each with concrete examples.
Declaration requirement: which assignments need an AI-use declaration, and how to write one.
Assessment adjustments: which assessments move to a process view, such as drafts, timed work, or an oral defence.
Handling suspected cases: the steps a teacher takes on suspecting AI use, education first, without convicting on a detection tool alone.
Review mechanism: the policy is reviewed once a year and updated as circumstances and the tools change.

Common questions

Should students be banned from using AI altogether?

A blanket ban is both impractical and against the direction of the programme. The programme itself asks schools to run AI-assisted teaching and to hold student AI literacy activities, which means teaching students to use it well. The workable approach is to divide by occasion: encourage it while learning and brainstorming, forbid it in assessments that measure a student's own ability. What students need to learn is the judgement to tell the two apart.

If a detection tool says the work was AI-generated, can we call it plagiarism outright?

We advise against it. Today's AI detectors get it wrong often, and convicting on a single score risks wronging a student who wrote the work themselves. A detection result is at most a starting point for a conversation with the student. The final judgement should draw on evidence from the process, such as draft records and how the student handles an oral defence.

Does this policy have anything to do with the EDB funding programme?

It does. The programme brings AI into the classroom across three subjects and asks schools to hold student AI literacy activities. A clear academic integrity policy can be where those activities land, and it can be written into the school's AI implementation plan as part of school-based governance.

Sources

EDB Circular Memorandum 221/2025 (Chinese)

Want to know your school's next step?

Tell us where your school is with the funding programme. We help with evidence, from choosing tools to filing the reports.

Book a free school consultation Request a proposal