Back to blog

AI in Education: 85% of Teachers Use It, Only 18% Were Ever Shown How

November 16, 2026 (3d ago) · 6 min read

AI in Education: 85% of Teachers Use It, Only 18% Were Ever Shown How

85% of educators use generative AI tools. Only 18% of US teachers have received any formal guidance on how to use them. That is the headline finding of Google's new AI in Education report, published this month.

The tools arrived in our classrooms years before the training did. The guidance is only now catching up.

What Google's report actually says

The report, AI in Education: From Potential to Practice, pulls together Google's own research, Gallup survey data, OECD PISA results, and published studies. The numbers that matter most for classroom teachers are these:

  • 85% of educators report using generative AI tools — but a Gallup survey found only 18% of US teachers have received any formal guidance from administrators on how to use them.
  • In higher education, only 6% of faculty strongly agree their institution provides sufficient training resources on using AI.
  • The policy gap is starting to close from the top: 40 US states and Puerto Rico now have official AI guidance frameworks for schools.

The report sets those figures against a decade of falling scores. Between 2015 and 2025, average OECD PISA scores fell by 28 points in reading and 22 points in mathematics — what the report calls "more than a full year of lost schooling." In the same period, global spending on education technology climbed from $107 billion in 2021 to $187 billion in 2025. We bought the tools. The results did not follow on their own.

The most quoted study in the report makes the point sharply. In a study published in PNAS, high school maths students with access to AI scored 17% lower on unassisted exams than peers who worked without it — but adding teacher-designed guardrails eliminated the drop entirely. The tool was not the problem. Unguided use was. Students sense it too: two-thirds of US middle school, high school and college students worry instant-answer tools degrade their critical thinking, and 60% worry classmates might misuse AI to gain an unfair advantage.

Why this matters more than the next new tool

Most of the debate about AI in schools is still framed as a fight about the tool: ban it, or embrace it. The report's numbers say the fight is in the wrong place. The difference between AI that helps and AI that harms is not the model — it is whether a teacher designed the way students use it.

Take Singapore. 66% of students there use AI chatbots weekly to support their learning, against an OECD average of 46%. Singapore leads the world in PISA reading and ranks second in maths and science. The difference is not that Singaporean students use AI more — it is that Singapore integrated AI literacy into the curriculum, training students to think critically about AI outputs and spot false information. Use plus guidance beats use alone.

The same pattern shows up on the teacher side. In a pilot in Northern Ireland, teachers using Gemini in Workspace saved up to 10 hours a week on routine admin — time they could spend with students instead. In Sierra Leone, teacher-led AI tutoring produced statistically significant learning gains. In both cases, the educator was in the lead.

The gap is practical, not bureaucratic: teachers need to know what to do on Monday morning when a student pastes an AI answer into their homework. A framework from the state helps. A routine in the classroom helps more.

In my classroom

I teach ICT, Computer Science and Robotics in Bangkok, and my students use AI every week — for debugging Scratch projects, for explaining Python errors, for brainstorming robot designs. I set the guardrails instead of fighting it.

My rule is simple and it has not changed since the first chatbot arrived in my classroom: the AI can help you think, but it cannot think for you. When a student uses AI to fix a bug in their code, they have to explain the fix back to me in their own words before the lesson moves on. If they cannot explain it, they have not learned it.

This term I ran that rule with my Year 7 and 8 students on a Scratch game project. The students who used AI as a tutor — asking it why their sprite would not move, then testing the answer themselves — finished with games they could explain and debug. The students who pasted whole scripts in without reading them got stuck the moment something broke, because the code was never theirs. Same tool, same classroom, opposite results.

Steal this: a five-step guidance gap fix

You do not need to wait for a national framework. Here is what I would do this week:

1. Write one page of class AI rules, with your students. Not a ban list — a use list. When AI is welcome, when it is not, and what "show your thinking" means in your subject. Students follow rules they helped write.

2. Make "explain it back" the default. Any AI-assisted work must be explainable by the student, in their own words, without the screen. This one routine kills most copy-paste cheating on its own.

3. Design one task AI cannot finish alone. A task that needs the student's own data, their own opinion, or a physical build — like a robotics prototype that has to actually move. If the answer can be pasted, the task is the problem, not the student.

4. Teach students to judge AI output, not just use it. Give students a deliberately flawed AI answer once a week and ask them to find the error. Spotting bad output is the AI literacy skill that matters most.

5. Reclaim one admin hour with AI. Pick one repetitive weekly task — worksheet variations, marking comments, lesson summaries — and let AI draft it while you check it. Use the hour you save for actual teaching.

The honest caveat

This is Google's own report, and Google sells the tools it praises. The headline figures come from real surveys — Gallup, OECD PISA, peer-reviewed studies — but the product recommendations are a vendor's view, not neutral advice.

The through-line is still worth acting on: 85% use it, 18% were shown how. That gap will not close itself, and it will not be closed by the next model release. It closes one classroom routine at a time — and teachers are the ones who close it.

Were you ever given formal guidance on using AI in your classroom — or did you have to figure it out alone like most of us?


Sources: Google, AI in Education: From Potential to Practice (November 2026) — full report (PDF).

Title image: original clay artwork created for this blog.

LINE