For school leaders
AI and academic integrity: what international schools can do
Most schools I work with started their response to AI by trying to catch it. That approach is running out of road. AI detectors aren't reliable enough to accuse a student, and in international schools the risk falls hardest on the students least able to argue back. One study found detectors flagged more than half of essays written by non-native English speakers as AI-generated (Liang et al., 2023). If a large share of your students write in English as an additional language, a detector score is a weak thing to stand on in a meeting with parents.
What the exam boards agree on
The IB, Cambridge International and JCQ (the body behind UK GCSEs and A-levels) word things differently, but five points come up again and again:
- The work has to be the student's own. AI output isn't the student's work, even when it's acknowledged.
- Any AI use has to be declared. Undeclared use is treated as misconduct.
- Teachers have to see the work develop. If a teacher isn't confident the work is the student's, it shouldn't be signed off.
- Software isn't the answer. The IB says detection software "should not be a substitution for responsible and purposeful teaching."
- When in doubt, supervision and conversation decide. Drafts under supervision, and a short discussion where the student explains their work.
The full board-by-board detail, including what each board asks students to record and where each requirement is met in a task, is in the Pendulum in Practice editions.
What this means for your school
Look at those five points together and every one of them is about how the work was made. A board wants to know that the thinking was the student's, and the only way to know that is to see it happen. So integrity can't rest on catching students afterwards. It has to be built into how tasks are designed, so the thinking leaves evidence along the way.
Where the Thinking Pendulum fits
The Thinking Pendulum gives your staff one shared way to describe how students think when they use AI: as a vending machine that hands over answers, or as a thinking partner that pushes their own thinking. Each part of the approach lines up with what the boards ask for:
- A stage-by-stage brief that says where AI is allowed and asks students to declare what it did (points 1 and 2).
- Process evidence such as a keep or reject log and a chat bibliography, so teachers can see the work develop (point 3).
- A rubric that marks the thinking, so decisions rest on evidence from the work itself (point 4).
- A short conversation about one decision in the work: why this choice and not another (point 5).
What changes across the school
Staff share one language, so a student hears the same expectations in History as in Biology, and departments put a share of the grade on the thinking as well as the product. For parents the message is simpler. The work their child hands in will show what their child can actually do.
Questions school leaders ask
Do we still need an AI policy?
Yes. A policy sets what's allowed. The Pendulum describes the thinking that happens inside those limits. If your school already uses an AI assessment scale, keep it. The two sit side by side.
What about the IB's changes for 2027?
The Extended Essay guide for first assessment in 2027 leans further towards process: the student's own work, AI acknowledged and cited where it's used, and the viva to confirm authenticity. Schools that already collect process evidence will find the change much easier.
Does this work without devices?
Yes. The evidence can come from paper, drafting in class and conversation. Sense checking, the first zone, doesn't need AI at all.