Last week, one of my AP World History students asked me something that stopped me mid-sentence. We were analyzing primary sources from the Columbian Exchange, and she’d been using Claude to help her understand a particularly dense 16th-century Spanish trade document. She looked up from her screen and said, “Mr. Jacobson, the AI gave me a really clean summary, but it completely missed the power dynamic between the merchant and the indigenous broker. Why would it do that?”
That’s the moment I live for. Not because the AI failed — it didn’t, really — but because my student had gotten good enough at thinking historically that she could see what was missing. She wasn’t outsourcing her thinking. She was sharpening it.
And that’s the story I keep coming back to as AI reshapes classrooms around the world: the tool isn’t the point. The thinking is the point. The question is whether we’re designing learning experiences that make that kind of thinking possible — or whether we’re just letting the machines do the heavy lifting while our students watch.
The Global Landscape: Everyone’s Building the Plane While Flying It
Let’s be honest — nobody has this figured out yet. But some fascinating experiments are happening, and many of the most interesting ones are outside the United States.
China has gone all in. Starting in fall 2025, Beijing became the first provincial-level region to introduce compulsory AI classes in public schools, with nearly 1,500 primary and secondary schools now offering AI instruction. Hangzhou followed suit, making AI courses mandatory across primary and secondary levels. The curriculum is structured and ambitious: third-graders learn AI fundamentals, fourth-graders tackle data and coding, and by fifth grade, students are working with intelligent agents and algorithms. Full nationwide rollout is planned by the end of this year.
What’s striking about China’s approach is the scale and intentionality. This isn’t a pilot program at a handful of elite schools — it’s a national strategy. AI has been labeled essential for national security and economic competitiveness, and the education system is being restructured accordingly. You can debate whether top-down mandates produce deep learning (I have my doubts), but the ambition is undeniable.
For those of us at international schools in China, this creates a fascinating tension. We’re watching our host country move faster than almost anyone on AI education policy while navigating our own institutional debates about how much AI is too much.
Inside China’s International Schools: Careful Steps, Real Conversations
A recent AmCham China report offers a revealing snapshot of how international schools across the country are handling the AI transition — and the picture is far from uniform.
Wellington College Tianjin is using adaptive platforms like Tassomai, IXL, and Kognity alongside generative tools like ChatGPT for lesson planning, and they’ve developed their own GenAI policy to guide responsible use. Western Academy of Beijing emphasizes transparency-focused AI use with tools like MagicSchool and Diffit, pairing them with staff workshops and ethics frameworks. QSI International School of Chengdu has built a Responsible AI in Learning (RAIL) framework that requires student training before they can access generative tools.
Meanwhile, Chengdu International School is deliberately avoiding AI-dependent assessments, emphasizing what they call “mastery and deep understanding” through critical thinking rather than technology shortcuts. And ISA Wuhan International is taking a cautious approach, prioritizing what their leadership calls “preserving the ‘human touch’ at the heart of instruction.”
The common thread? Every school I’ve read about or talked to is grappling with the same core worry: metacognitive laziness — students outsourcing their thinking to AI rather than engaging with the material themselves. Schools are redesigning assignments to require more in-class work, transparent AI citations, and genuine intellectual struggle. That last part matters more than people realize.
The Critical Thinking Paradox
Here’s the tension that keeps me up at night: students are using AI more than ever, and they’re increasingly worried it’s making them worse thinkers.
According to EdWeek’s March 2026 survey, concern among middle schoolers jumped from 48% to 68% in less than a year. High schoolers went from 55% to 65%. College students? Seventy percent now believe AI use may be harming their critical thinking. And yet usage keeps climbing — 46% of middle schoolers and 60% of high schoolers now use AI for homework regularly.
Heather Schwartz from RAND put it well: “That initial struggle, the blank page… that is like a tiny moment of friction. It’s a cognitive challenge… AI might be giving you really beautiful explanation… It’s still removing that step for you.”
That “tiny moment of friction” is everything. In my AP World classroom, I’ve started calling it “productive confusion” — that stretch where a student doesn’t quite understand a source, doesn’t quite see the connection, and has to sit with the discomfort long enough for real understanding to emerge. AI can short-circuit that process before it even begins.
But here’s what I’ve also noticed: when students learn to use AI after they’ve done the hard thinking — as a check, a sparring partner, a way to stress-test their own analysis — it actually deepens their understanding. The tool amplifies whatever cognitive habits are already in place. If those habits are strong, AI makes them stronger. If they’re weak, AI makes them weaker.
Media Literacy Is the New Battleground
Colleen Kenny makes a compelling argument in her recent LinkedIn piece that media literacy has become “democracy’s most urgent task.” Her central insight is worth sitting with: “What we desperately need is not adherence to a single dominant narrative, but rather shared methods for establishing what’s true.”
She’s right, and the implications for our classrooms are enormous. Trust in media has plummeted to 28% in the U.S., with similar declines across democracies worldwide. When AI can generate convincing video, realistic images, and polished prose indistinguishable from human-created content, the old skills of “check the source” aren’t enough anymore. Students need multilayered literacies — cognitive, emotional, and embodied — to navigate what Kenny calls “content chaos.”
This is exactly where history teachers have an edge. We’ve always taught source analysis, perspective-taking, and the interrogation of evidence. The difference now is that our students need those skills not just for understanding the past, but for surviving the present.
The policy world is catching up. Media Literacy Now reports that more than half of U.S. states have now taken legislative action on media literacy, with eleven enacting new laws since January 2024. And the integration of AI literacy with media literacy is increasingly recognized as inseparable — you can’t understand modern media without understanding the AI systems that create and curate it.
What I’m Doing in My Classroom
I don’t pretend to have cracked this, but here’s what’s working for me right now:
Source Analysis with AI as a Second Opinion. When students analyze a primary source — say, a passage from Ibn Battuta or a Qing dynasty edict — they do the initial analysis by hand. Then they ask Claude to analyze the same source and compare interpretations. The conversation that follows is almost always richer than either analysis alone. Students get genuinely excited when they catch something the AI missed, and they get genuinely curious when the AI surfaces a connection they hadn’t considered.
“Interrogate the Output” Exercises. I regularly have students generate AI responses to historical questions, then systematically evaluate those responses for accuracy, bias, missing context, and oversimplification. It’s a media literacy exercise disguised as a history lesson, and it builds exactly the kind of critical evaluation skills that transfer to evaluating any information source.
Transparent AI Use Policies. Students in my class can use AI, but they have to cite it — what tool, what prompt, and what they did with the output. The goal isn’t policing; it’s metacognition. When you have to articulate how you used a tool, you think more carefully about whether the tool was actually helping you think or just helping you avoid thinking.
The Bigger Picture
A forum on AI in China’s education system surfaced a quote from an art teacher at Suzhou Experimental Primary School that I haven’t been able to shake: “AI can create beautiful paintings, but it can’t read the crooked little happiness in the sun painted by children.”
That’s it. That’s the whole thing.
AI can produce polished essays, generate lesson plans, summarize dense documents, and create beautiful artwork. But it can’t see the crooked little happiness — the messy, human, deeply personal meaning that students bring to their work when they’re truly engaged. Our job as teachers isn’t to compete with AI’s polish. It’s to protect and cultivate that irreplaceable human element.
Anthropic’s partnership with Teach For All to offer AI tools and training to educators in 63 countries, reaching over 100,000 teachers, is a step in the right direction. Purdue University requiring AI working competency for all undergraduates starting fall 2026 signals that higher education is taking this seriously too. The 400,000-teacher training initiative from the American Federation of Teachers shows the scale of what’s needed.
But here’s what I keep coming back to: none of these initiatives matter if teachers aren’t at the center of the conversation. As the educators at that Suzhou forum put it, “Teachers, not technology, should lead educational transformation.” The technology is moving fast. Our students need us to move thoughtfully.
David Jacobson is an AP World History teacher at an international school in Shanghai. He writes about AI and education at shouldiuse.ai. You can find him on LinkedIn or reach him at dawidio@gmail.com.
