A generation ago, ‘looking something up’ meant finding a fact or reading what someone else had said about a topic, likely in a different context to the one you were considering. The default medium evolved from a bookshelf to a CD-ROM and then a web browser during my own school and university days, but the fundamentals never changed. We would find the information we needed, in whatever form it was available, then adapt it to answer the question we had been set.

Countless articles, studies and books have been written on the topic of assessment in schools, generally analysing specific methodologies or discussing the very purpose or validity of assessing students. Regardless of the ideology or motivation, I think it’s reasonable to boil it down to a single core principle: the need to gauge students’ understanding and ability to apply their knowledge. Before the AI age, it was relatively straightforward to design a task that created a measurable gap between the raw materials available to a student and the output expected from them. The extent to which they were able to close it could then serve as a reasonable proxy for their understanding of the topic.

What’s different now is that information retrieval itself can be highly contextual: tailored not just to the topic, but to the specific level and question being asked. A student today can prompt their way to a complete, reasoned position on a genuinely contested question, argued from both sides and expressed with whatever level of sophistication they wish. When there is no difference between the material the student can extract from an external source and the output they are expected to produce, the assessment proxy disappears. No matter how well the rubric is designed, it is ultimately not measuring the student’s understanding and therefore is not serving its core purpose.

Stop calling it cheating

Students ‘outsourcing’ their work is a tradition as old as education itself. In my day, those who either couldn’t or didn’t want to do the work would pay an older or more talented peer, or a tutor, to do it. Younger millennials and early Gen Zers didn’t need to meet the person at all; they could pay a stranger on the internet instead. Still, there was a financial and potentially social cost that meant plagiarists remained a small minority. An eagle-eyed teacher who knew a student’s usual standard of work could generally tell when a piece didn’t match it, and the prevalence made the problem manageable so long as the school had a decent academic integrity policy.

According to the College Board, 84% of US high school students used generative AI for schoolwork in 2025¹. Anecdotally, I have observed and been told of similar patterns in UK and international schools. This isn’t a case of spotting one or two cheats any more. In fact, I think we need to stop thinking of it as cheating at all. Rather, I support the increasingly popular view that there is a need to rethink the very nature of assessment, accepting AI as part of how students work rather than trying to minimise it or pretending it isn’t happening.

Universities are moving faster on this than schools are. A 2026 study of over 95,000 students across twenty major US research universities, published in Science, found substantial variation in how and how much students use generative AI by discipline. It argued for assessment reform tailored to each subject rather than blanket bans or reliance on detection software². Vanderbilt disabled its AI-detection software back in 2023, after calculating the tool’s own claimed error rate would have wrongly flagged around 750 genuine student papers a year, and dozens of other universities have since done the same. Schools have been slower to follow. Many are still relying on detection software of exactly the kind those universities gave up on, without the same scrutiny of whether it actually works.

An opportunity, not a threat

This is the best opportunity schools have had in years to step up what they ask of students when it comes to intellectual engagement. Content is not losing its place at the centre of education. What’s gone is the excuse for treating its slow accumulation as a precondition for anything else.

For students in all but the most elite schools, a certain type of work has always waited until late secondary education or even university: building a position across sources that disagree, then defending it under real challenge. In many cases, the rationale has been one of intellectual readiness: younger students are still developing the skills to have that kind of discussion. Having seen middle school students out-debate competitors in the final year of their IB Diploma on more than one occasion, I can attest to the fact that this is simply not the case.

I believe the real reasons for the historical pace of development have been twofold. First, building and defending a position needs enough ammunition to argue with. Assembling that much material in one’s mind, let alone synthesising it properly, used to take years. Then there is the systemic challenge: a premise that cascades down to younger students, with phrases like ‘that’s a year 9 topic’ regularly parroted to curious but chronologically challenged minds. Centrally planned and universally delivered mass education has long been reliant on progression through the content at something close to a fixed annual rate.

The first of those two problems has collapsed. A thirteen-year-old with internet access can now assemble in an afternoon an argument that would once have taken even an undergrad weeks or months to gather. That doesn’t hand them a scholar’s judgement, but it does remove the honest reason for making them wait years before they start building it. Few, if any, established curricula are designed around the level at which a student of a given age can think. Most are built as spirals, designed to cover increasingly challenging or complex content over time, regardless of what any individual student might already be capable of. Untangling a curriculum built around a fixed pace, and trusting a younger student with work a school has always reserved for someone older, takes a kind of institutional bravery most schools and systems haven’t yet found.

Change is afoot

Whilst change may be slow, as it always has been, some programmes are already starting to respond. The International Baccalaureate’s new Systems Transformation course³ replaces two full standard-level subjects with three hundred hours of collaborative work, built for students to take informed, ethical action on real, contested problems. One pilot cohort spent theirs converting food waste into biogas for communities in Kinshasa. That’s closer to a university capstone than a school assignment, attempted by students who haven’t sat a final school exam yet.

What makes this one worth taking seriously is that losing two subjects hasn’t meant losing intellectual challenge. If anything, what’s replaced them asks more of students, not less. The IB’s own language for its assessment model is explicit: it wants to test “not only what students know, but how they apply knowledge”. Two ‘traditional’ subjects’ worth of content acquisition, gone. In its place, exactly the kind of work students used to only come across through a gifted and talented programme, at university, or in the workplace.

It isn’t proven yet. Three cohorts in, and universities are still being persuaded the rigour survived the subtraction. It’s the honest state of something still being tested in public, which is further than most curriculum reform ever gets. To me, it feels like a positive direction: an acknowledgement that this change isn’t just inevitable, it’s already happening.

Time waits for no man

A survey of more than 1,000 US employers, released by the American Association of Colleges and Universities in December 2025, found that applying knowledge to real-world problems was their single most valued skill in graduates, ahead of subject knowledge itself, with critical thinking and judgement close behind⁴. Fields now move fast enough that a consensus answer can be out of date before the course that taught it has finished. Recall is worth less in that environment. What holds its value is reaching and adapting a defensible position while the ground is still moving, which is a different, harder thing than reciting where the ground used to be.

Systems Transformation proves this is achievable at the oldest end of school, where the institutional bravery required is smallest. The harder version, reaching all the way down to a thirteen-year-old with internet access and a curious mind, is still mostly unproven. That’s not a reason to wait for someone else to prove it first.

The bar set by current assessment schemes is easier to clear than it ever has been, whether you call the method ‘cheating’ or ‘making effective use of available tools’. Our response should be to raise it, not to trip students on the run-up.

¹ College Board, “New Research: Majority of High School Students Use Generative AI for Schoolwork,” October 2025: https://newsroom.collegeboard.org/new-research-majority-high-school-students-use-generative-ai-schoolwork

² Igor Chirikov, Ivan Smirnov, René F. Kizilcec, “Generative AI use and misuse call for assessment reform in higher education,” Science 392, 818–820 (2026): https://www.science.org/doi/10.1126/science.aec5115

³ International Baccalaureate, “IB to expand new Systems Transformation course following Board approval,” May 2026: https://ibo.org/news/news-about-ib-schools/international-baccalaureate-to-expand-new-systems-transformation-course-following-board-approval/

⁴ American Association of Colleges and Universities and Morning Consult, employer survey, reported in Natalie Schwartz, “7 in 10 employers have high confidence in higher ed, survey finds,” Higher Ed Dive, December 2025: https://www.highereddive.com/news/7-in-10-employers-have-high-confidence-in-higher-ed-survey-finds/807607/