On this page
How the sector arrived at 2026
The first wave of school responses to ChatGPT in early 2023 was almost universally restrictive. Several international schools issued blanket bans, blocked the relevant URLs on school networks, and treated any suspected use as a disciplinary matter. Within a year, most had reversed course. The reasons were practical. The bans were unenforceable on personal devices and home networks; the supposed detectors were unreliable; the examination boards declined to treat all AI use as malpractice; and the parent body began to ask a more uncomfortable question, namely whether their children were being prepared for a world in which AI fluency would be ordinary.
By the start of the 2024 academic year, the dominant model had shifted to conditional permission. Schools allowed AI for ideation, drafting and feedback in classroom contexts, prohibited it for unsupervised submitted work, and required disclosure when used. A second wave of policy revision through 2025 added more granularity: differentiated rules by year group, defined rubrics for what counted as appropriate use within each subject, and the first wave of internal AI literacy programmes built into the curriculum.
The 2026 snapshot is the result of three years of iteration. The schools that have invested have built coherent policies, trained their staff, and integrated AI use into assessment design rather than fighting against it. The schools that have not invested have policies on their websites that bear little resemblance to what happens in classrooms. The gap between the two groups is now wider than the gap between any two other dimensions of school quality.
The four policy patterns
The international sector in 2026 falls into four broad policy patterns. Recognising which pattern a school sits in is the most useful early filter you can apply.
Free download: AI policy comparison checklist
Our one page checklist covers the eight questions to ask any school about AI policy, with annotated examples from schools across the four patterns. Subscribe below to receive it in the welcome email, or send a request through the Get Help form. For a wider view of the school year ahead, see our State of International Schools 2026 annual report.
What changed in assessment
The biggest practical changes in the past three years have been in assessment design. The IB, Cambridge and AP boards have all maintained closed book external examination as the dominant assessment for senior students, which has insulated the headline grades from AI disruption. Where AI has changed school practice is in coursework, internal assessment, extended essays and the day to day rhythm of homework.
The most visible change has been the rise of supervised in class composition. Tasks that were previously homework essays are now timed in class writing, often handwritten, with the students given the prompt at the start of the lesson. Strong schools have made this shift deliberately and explained it to families; weaker schools have made it informally, leaving parents unsure why the homework volume has dropped.
The second change has been oral defence of submitted work. Students are increasingly required to discuss their submitted essay or coursework in a one on one conversation with the teacher, who probes whether the student can defend the argument and explain the choices they made. This format borrows from university viva traditions and is likely to spread further. The IB extended essay, the EPQ in the British system and AP Capstone all now include oral elements that were either added or strengthened in response to AI.
The third change has been the rise of process artefacts. Students submit not just the final piece of work but the drafts, the prompts they used with AI tools, the research notes and the changes they made between versions. This documentation matters more than the final product in some assessment regimes. Schools that have built the workflow into the way students use word processing tools have made this easier; schools that have left it to the student tend to find the documentation patchy.
For the underlying curriculum trade off, our curriculum hub sets out the structural details, and our IB versus AP university outcomes piece compares how the two systems have responded.
AI literacy as a curriculum strand
The schools that have moved furthest have built AI literacy into the curriculum as a strand rather than a one off lesson. The strand typically covers four areas: how the underlying models work at a level appropriate to the year group, what AI is and is not good at, the ethics and privacy considerations, and the practical skills of writing useful prompts and evaluating outputs critically. The strongest programmes integrate this work into existing subjects rather than creating a separate AI lesson, on the principle that AI fluency is a method of working that should be present in everything.
In primary, the focus tends to be on critical evaluation of AI outputs, with children given AI generated text containing errors and asked to identify them. In lower secondary, students learn the structure of prompting, the limits of what models can do reliably, and the basic ethics of attribution. In sixth form, the work shifts to using AI as a research and feedback assistant within the discipline, while preserving the student's own argument. This progression is increasingly visible in the better schools.
Whether your child's school has this strand is one of the most useful questions you can ask. A school that cannot tell you what is taught about AI in Year 5, Year 8 and Year 11, with examples, has not built the curriculum strand. The marketing language about being AI ready is widespread; the substance is concentrated in a smaller number of schools.
Specific examples from schools
To give a sense of what is being built where, several international schools have published their approach in detail. Stamford American International School in Singapore runs a clearly described conditional permission model with disclosure rubrics. UWCSEA in Singapore has built a school deployment of Microsoft Copilot for senior students, with logged use and integration into the IB programme. Brighton College Dubai has published a graduated approach across year groups with explicit examples of acceptable and unacceptable use.
In London, the major international schools (ACS, ASL, Southbank International) have all moved toward conditional permission models, with the more academically selective British independents leaning further toward supervised in class composition for examination subjects. In Hong Kong, the ESF system has standardised an approach across the family of schools, which gives families consistent expectations across primary and secondary placements.
The pattern across these examples is convergence. Three years ago, school AI policies looked nothing like each other. By 2026, the strongest schools are arriving at very similar positions: open use with disclosure, integrated literacy programme, supervised composition for high stakes assessment, oral defence for major coursework. This convergence is itself useful information for families; a school whose policy looks markedly different from this consensus is either further ahead than the rest of the sector or has not yet engaged.
Questions to ask in admissions
Most admissions teams have prepared answers to the open ended question "what is your AI policy." Specific questions produce specific answers and a more useful picture.
Could you describe what is taught about AI in Year 5, Year 8 and Year 11, with one specific example from each? A confident school answers with examples. A weaker school describes intentions.
How is the policy different across subjects? The strongest schools have differentiated rules; the weakest have a single sentence applied to everything.
How has assessment changed in the past two years in response to AI? Look for specifics: timed in class composition, oral defences, process artefacts. Vague answers usually mean little has changed.
What AI tools does the school provide centrally, and which are students expected to access on their own? A school deploying its own AI environment is a different proposition from one expecting families to provide ChatGPT subscriptions.
How does the school handle suspected AI misuse, and what was the most recent example? The strongest answers describe a defined process with proportionate consequences. The weakest answers describe zero tolerance with vague enforcement.
For the broader question set across all admissions visits, our 10 questions every parent should ask before choosing a school piece sits alongside the AI specific questions above.
What is likely to change next
Three changes are likely in the next two academic years. First, examination boards will continue to add AI specific guidance to coursework rubrics, and the IB in particular is expected to formalise its expectations around AI use in extended essays and internal assessments. Schools that have not built the workflow for documenting AI use will find themselves catching up.
Second, the gap between schools that deploy their own AI environments and schools that rely on consumer tools will widen. The data privacy conversation is now part of every school technology decision; expect more schools to move to closed sandbox models for senior students, particularly in jurisdictions with strong data protection regimes.
Third, AI literacy will start to appear in school comparison frameworks. The way that ESL provision became a comparison axis a decade ago, and SEN provision more recently, AI literacy will become a question parents ask routinely. Schools that prepare for this question will benefit; schools that do not will be left explaining why the answer is thin.
For the wider trends shaping the sector, our cyber safety standards piece (where it sits in the trends cluster) covers the related digital safety questions, and our hybrid and online international schools piece covers the format question that often accompanies AI discussions.
Parent AI conversation checklist
- Read the school's published AI policy in full
- Ask how the policy varies by year group and subject
- Ask for one specific example of AI literacy taught in each key stage
- Ask how assessment has changed in the past two years
- Ask whether the school provides AI tools centrally or expects families to do so
- Ask how suspected AI misuse is handled, and for a recent example
- Discuss the policy with your child so they know the rules
- Set household norms for AI use during homework
FAQ
Most do, but with conditions. The dominant pattern is allowed for ideation, drafting and feedback in lessons, prohibited for unsupervised submitted work, and required disclosure when used. A small minority have moved to closed AI environments where students use a school controlled model rather than the public ChatGPT.
The IB, Cambridge and AP boards have all maintained closed book examination as the dominant assessment, with AI generated text not permitted in coursework or extended essays. The biggest changes have been in coursework verification, oral defence of submitted work and supervised in class composition. Schools that ignore these shifts will leave students unprepared.
Worry less about cheating, more about whether the school is teaching students to use AI well. Schools with no AI literacy programme are leaving students with neither the discipline to write without AI nor the skill to use it productively, which is the worst of both worlds.
Most schools introduce AI literacy concepts from late primary (Year 5 or 6) and tool based work from lower secondary onwards. The exact starting point varies; the more important factor is whether the introduction is structured and progressive across year groups.