The class of 2026 is the first cohort of students who completed high school with full access to AI writing tools. Sixty-seven percent of them now say using AI for schoolwork is hurting their own critical thinking, up ten percentage points in ten months [1]. That's the students' own assessment of what AI is doing to them.
Research published in late 2025 found a strong statistical link between frequent AI tool use and declining scores on standardized critical thinking assessments, in a controlled study of 666 participants [2]. The numbers from faculty are just as stark. A survey of 1,057 college and university faculty conducted in October-November 2025 found 90% said generative AI will diminish students' critical thinking skills, with 66% believing the impact will be substantial [3]. In April 2026, Minding the Campus summarized early data on the class of 2026: sharp declines in writing quality and the ability to construct a sustained argument, in the first cohort of students who had AI access throughout their diploma years [4].
Meanwhile, usage kept climbing. Among UK university students, AI use jumped from 66% to 92% in a single year, according to a 2025 HEPI survey [5]. Patterns at the K-12 level followed the same direction: RAND found middle and high school AI use for homework rose from 48% to 62% between May and December 2025 alone [1]. Around 40% of schools had outright bans in place during that period, which means much of the increase was invisible to policy. Students used AI at home, submitted the output at school, and teachers worked with incomplete information about what they were actually assessing.
What Makes This Harder in International Schools
International school teachers face this without the guidance infrastructure that domestic teachers get. In the US, 34 states have issued formal AI guidance to districts, with Ohio, Virginia, and Tennessee among those mandating district-level policies [8]. International school teachers largely make individual classroom calls, sometimes with a school policy, often without one, almost always without formal training on the subject.
About 45% of educators globally say they haven't received any AI training at all [6]. In international schools, where professional development typically ties to IB and Cambridge certification priorities rather than tech policy, the training gap is likely wider still.
The result is teachers making consequential classroom decisions based on personal judgment, student behavior they've noticed, and informal conversations with colleagues. Some administrations pushed AI integration hard in 2023 and 2024, without specifying what that should look like in practice. A few are now revising that position. Nobody's announcing it publicly.
The IB Angle
The IB officially permits AI use and has published principles for AI integration, making clear that AI-assisted work isn't automatically academic misconduct [9]. The practical implication for teachers: assessment design needs to account for the possibility that AI did some of the work, which is a very different question from whether a student plagiarized. Most IB teachers were left to figure out what that meant for their classrooms individually. Some are still working on it.
Internal Assessments for the class of 2026 have come under tightened scrutiny. This is the first full IB diploma cohort who had ready AI access across all three years of their programme. Teachers who've been writing IA tasks the same way for a decade are finding those tasks don't hold up as evidence of student thinking the way they used to. The IBO's guidance acknowledges this. The professional development to support the adjustment hasn't materialized at the same speed.
The Ban or Integrate Debate
The schools responding most visibly have split. Ban AI use entirely, or build it into curriculum. Around 40% were in the ban category as of 2025 [6].
The case for banning is practical enough. Student AI use at home makes enforcement impractical in one sense, but classroom bans create observable conditions: in-class writing, oral components, tasks students can't complete by pasting into a chatbox. Teachers who've moved in this direction often report being able to assess student thinking again, which had become genuinely harder to do.
The case against banning runs: students will use AI after they graduate, teaching them to use it critically is more useful than pretending it doesn't exist, and schools that ban it may leave students unprepared for workplaces where AI is standard. Brookings reviewed the current evidence in early 2026 and concluded that "the risks are overshadowing the benefits" in how generative AI is being used in most school contexts today [7]. That's not a ringing endorsement of either camp. It does suggest the debate won't resolve cleanly.
For international schools specifically, the ban question has an additional wrinkle. Students with home access to premium AI tools will use them regardless of what the school network allows. A classroom policy that bans AI doesn't prevent use. It changes the visibility of it.
What's Actually Working
The teachers getting the most reliable results in 2026 aren't uniformly on one side of the ban debate. The pattern in classrooms where student learning seems to be holding up is redesigned assessment rather than better AI policy.
That looks like oral defenses of written work, in-class writing under observation, portfolio tasks where the process across drafts is visible, assignments that require students to explain their thinking rather than produce a final product. None of this is new pedagogy. It existed before AI, and AI made it more necessary by making the old formats easy to fake.
The more useful classroom conversation, and the one some teachers have started having explicitly, is teaching students what AI doesn't do. A language model produces fluent text with no comprehension behind it. Students who understand this clearly and can articulate the difference between AI output and their own thinking have something more useful than a policy: a reason to do the work themselves. Several teachers who've tried this approach report it shifts motivation in ways that ban enforcement hasn't.
Conclusion
In the schools handling this best, individual teachers changed their assessment design before any formal AI policy existed. If your school is still in the ban-vs-integrate debate without asking what students are supposed to be demonstrating, the debate is probably happening at the wrong level. The students arriving in September have been using AI tools for three years. They're already ahead of most policy conversations.
