What the IB requires, and what the policy is for
An academic integrity policy is one of the written policies an IB World School is required to have. It is requested at authorisation, revisited at programme evaluation, and — unlike most policy documents — it gets used, because a breach forces somebody to open it.
The IB's own academic integrity documentation and the assessment procedures for each programme are the authority on definitions, obligations and the consequences the IB itself applies. A school policy sits underneath those and does a different job: it says what this school teaches, expects, and does.
The test of the document is not whether it defines plagiarism. It is whether a teacher who suspects something at nine on a Tuesday morning can find out what to do next.
The behaviours, defined so a fifteen-year-old understands them
The IB frames academic misconduct as behaviour, intentional or not, that gives or could give an unfair advantage in assessment. The 'or not' matters: a student can commit a breach through carelessness, and a policy that assumes intent will not describe most of the cases a school actually handles.
Plagiarism is using another person's ideas, words or work without properly acknowledging them. Collusion is supporting another student's misconduct, typically by letting them copy work. Duplication is submitting the same work for more than one assessment component. And there is a broader category of misconduct covering unauthorised materials or communication in an assessment, and presenting generated content as one's own.
Write each with an example a student in your school would recognise. Definitions in the abstract are the reason students say they did not know.
- Plagiarism — words, ideas, images, data and code, not only prose
- Collusion — including allowing a friend to see finished work
- Duplication — the same piece submitted twice, across subjects or components
- Other misconduct — unauthorised materials, communication, and unacknowledged generated content
The AI section, which is now the hard part
Presenting AI-generated content as your own, or failing to acknowledge meaningful AI assistance, is a breach. That is the easy half. The hard half is that AI is also a legitimate tool, students will use it, and a policy that bans it outright will be ignored and will teach nothing.
The policies that are holding up do three things. They distinguish between using a tool and submitting its output. They require acknowledgement in a specific, stated form rather than in principle. And they set expectations per task rather than per school, because the honest answer differs between a language acquisition writing task and a science internal assessment.
The mechanism that works is a per-task statement. Every assessed task says what assistance is permitted for it, and the default when nothing is said is that ordinary academic conventions apply. That shifts the conversation from a policy argument to a task instruction, which is where it is actually manageable.
Writing the AI section so it survives next year
The main drafting risk is naming tools. A policy listing specific products will be out of date within the year and will imply that anything unnamed is allowed.
Write about categories of assistance instead: generating text or code, restructuring the student's own writing, translating, summarising sources, generating images or data. Those categories will outlive any product. Say what acknowledgement looks like, give the form, and require it in the same place every time so a teacher can look for it.
Be equally explicit about detection, because this is where schools overreach. AI detection tools are unreliable and produce false positives, disproportionately against students writing in an additional language. A policy that treats a detector score as proof will eventually accuse a student who did nothing, and that is a far worse outcome than a missed breach. Detection can raise a question; it cannot answer it.
Who is responsible for what
A policy that assigns all responsibility to students is a policy about punishment. Spread it explicitly.
Students are responsible for the authenticity of what they submit and for acknowledging assistance. Teachers are responsible for teaching referencing within their subject, setting tasks that make wholesale substitution hard, stating what is permitted per task, and raising concerns rather than resolving them privately. Coordinators are responsible for consistency across subjects and for the school's dealings with the IB. School leaders are responsible for the policy existing, being taught and being reviewed. Families are responsible for understanding what help at home is appropriate.
That last one is worth writing down. A great deal of what schools handle as collusion is a well-meaning parent.
What happens when a breach is suspected
This is the section teachers will actually read, and it should be a procedure rather than a principle.
State who the concern goes to and within what time. State that the student is spoken to and heard before any conclusion, and who is present. State what evidence is gathered and by whom. State who decides. State what happens to the mark in the meantime. State how it is recorded and who is told, including whether and when the family is informed. And state what happens for a first instance versus a repeat, because consistency across two teachers is the thing students and parents notice.
For work that goes to the IB, the school's procedure sits alongside the IB's own, and the IB's published procedures are the authority on what must be reported and what it does about it.
Consequences, and keeping them proportionate
A policy with one consequence for every breach will not be applied, because teachers will quietly decline to trigger it for a minor case.
Build a range. A first, clearly careless referencing failure in a formative task is a teaching moment with a resubmission. A deliberate substitution in a piece of assessed coursework is a different matter with a different process. Say which is which and who decides where a case sits. Proportionality is what keeps a policy in use rather than in a drawer.
Teaching integrity rather than policing it
The schools with the fewest breaches are not the ones with the strictest policies. They are the ones where referencing is taught inside subjects, repeatedly, and where tasks are designed so that a student's own thinking is visible in the process rather than only in the product.
Process evidence is the most underrated defence available. A task with a proposal, a draft, a supervisor conversation and a final piece is very hard to substitute wholesale, and it also produces exactly the trail a school needs if a question is later raised. The extended essay and the personal project already work this way, and the same idea transfers down into ordinary coursework.
Recording, and what a school can actually show
This section names our own product, so read it as that. Everything above is true whatever systems you use.
When a breach is alleged, a school's position depends entirely on what it can show: what the student submitted and when, what drafts preceded it, what the supervisor said at each stage, and who made which decision. Schools that keep this across email, a drive and somebody's memory find out how thin that is at exactly the wrong moment.
skodle holds that trail as records. Work submitted by a student sits against the task it belongs to, with timestamps. Supervised components — the extended essay, the personal project, internal assessment — run through milestones with drafts and supervisor feedback attached, so the process evidence exists as a by-product of supervision rather than as something reconstructed afterwards. Every change carries an audit entry naming the person who made it.
The AI side is governed rather than assumed. High-stakes drafting in assessment and accreditation requires a coordinator to approve before anything is acted on, and every AI run is logged with the model, the prompt version and what it cost, so a school can answer what the AI did and who signed it off. That is a statement about staff use of AI inside the platform; it is not a detector, and skodle does not claim to identify AI-generated student work. Given how unreliable detection is, that is deliberate.
Review, and the date on the front
Name an owner, usually the programme coordinator with the head of school, and commit to a review interval. Given how fast the AI question is moving, an annual review is the minimum and a mid-year check on the AI section is sensible.
Review it with teachers and students in the room. A policy written in an office describes a school nobody in the building recognises, and students who helped write the AI section are markedly more likely to follow it.
The mistakes worth avoiding
Four recur. Banning AI outright, which drives use underground and forfeits the chance to teach judgement. Treating detector output as proof, which will eventually accuse an innocent student. Writing definitions with no examples, which hands every student the defence that they did not know. And a single consequence for every breach, which guarantees the policy is bypassed in the small cases and therefore inconsistent in the large ones.
The version that works is short, has examples, names who does what, offers a proportionate range, and is reviewed while the technology is still moving.
