The IB chose not to ban AI, and required it to be cited

The classroom note

On 1 March 2023, the International Baccalaureate published a statement about ChatGPT that a Diploma Programme teacher could hand straight to a class. It said plainly that 'the IB will not ban the use of AI software', calling a ban 'an ineffective way to deal with innovation'. For a teacher marking an extended essay, the instruction was specific: any text, image or graph produced by an AI tool and included in a piece of work 'has been copied from such software' in the same sense a quotation is copied, so it 'must be credited in the body of the text and appropriately referenced in the bibliography'. An essay that is predominantly quoted material, the statement adds, 'will not get many, if any, marks' under an IB mark scheme, whether the quoted material came from a book or a chatbot.

What the evidence says

The statement is the IB's own account of its position, not a third party's summary, and it is explicit about the underlying principle rather than a detection method: work produced 'even only in part' by an AI tool is not the student's own unless declared. The IB's separate academic integrity page, current as retrieved on 16 September 2026, reinforces the same framing without naming AI specifically: it asks students to be 'content creators', not 'content imitators', and to 'give credit where credit is due'. Neither document reports a study, a detection rate or a measured rule-breaking figure; both are policy statements describing what the organisation expects, not evidence of how consistently that is followed across thousands of IB World Schools.

The implementation question

The policy puts the referencing burden on the student and the judgement burden on the marker. A citation standard built for quotations and paraphrase has to absorb a new kind of source: a tool that can produce an entire draft, with no fixed edition or page to point to. The statement specifies no required citation format for an AI tool, leaving schools to adapt existing conventions, and describes no AI-detection technology the IB itself uses; enforcement rests on markers noticing an essay that reads as mostly quoted, unoriginal material, however it was produced. That is a judgement about voice and originality, not a technical scan.

What holds and what fails

What holds is the underlying logic: treating AI output like any other borrowed material, rather than banning a tool a mark scheme cannot see, keeps the rule stable as the specific software changes. What is likely to fail, editorially, is an assumption that a citation requirement is self-enforcing; a student who omits the credit has, by the IB's own description, committed exactly the malpractice the statement warns against, and the policy depends on a school culture willing to discuss what a citation-worthy use of AI looks like, a training question the statement leaves to 'schools to discuss the various types of academic misconduct with their students'.

• Would a marker here recognise an uncredited AI-generated passage as readily as an uncredited quotation from a book?
• What citation format has this school actually agreed on for AI tools, and is it written down for students?
• Does 'predominantly quotes' get judged the same way whether the quoted material is AI-generated or conventionally sourced?

The IB's choice to extend an existing citation principle, rather than write a new rule for a new technology, is a defensible design decision. Whether it works in a given classroom depends less on the statement itself than on whether that classroom ever discussed what counts as a citable use of a chatbot.