How AI Usage Is Growing Among Students and Teachers in 2026: What the New Survey Data Actually Shows

A data-driven look at how classrooms, homework, and grading changed this year — pulled from the newest higher-ed, K-12, and teen surveys, not recycled 2024 numbers.

By Oyekale Olawale · Updated August 2026 · 9 min read

I’ve spent the last two years watching this shift happen from both sides of the screen — testing the same tools students and teachers are now using daily, and digging through every major survey that tries to measure it. What struck me putting this update together isn’t that AI usage grew. Everyone expected that. It’s how unevenly it grew, and how far institutional policy has fallen behind actual classroom behavior.

This is a full rewrite of our earlier piece on this topic, built from primary survey data published in 2025 and 2026 rather than aggregated stat roundups. Every figure below is traceable to its original source. Let’s get into what actually changed.

The Headline Numbers: Adoption Jumped Faster Than Anyone Predicted

The clearest signal of 2026 comes from the Digital Education Council’s AI in Higher Education Global Survey 2026, one of the largest datasets ever assembled on this topic — 45,398 responses from 27,284 students and 18,114 faculty members across 35 countries. It found that 88% of students now use AI in their learning and 77% of faculty use it in their teaching, a 16-point jump on the 2025 numbers. That’s not incremental growth. That’s a technology going from “common” to “assumed default” in twelve months.

The K-12 picture tells a similar story, just on a different curve. The Gallup–Walton Family Foundation “Teaching for Tomorrow” study, run through the RAND American Teacher Panel with 2,232 U.S. public school teachers, found 60% of teachers use AI in some form — up from roughly a third of teachers just two years earlier. Only 32% use it weekly, but that group reports saving 5.9 hours per week, which the researchers translate into roughly six reclaimed weeks across a school year.

Where the Growth Is Actually Happening

Look past the headline percentage and the interesting part is what students and teachers are actually doing with these tools. On the student side, the DEC survey found 61% say AI frees them to focus more on thinking through ideas rather than mechanical tasks, and 31% report attempting more challenging work than they would have otherwise. That’s a genuinely positive signal — students aren’t just outsourcing thinking, many are using the freed-up time to go deeper.

But the same survey also surfaced something researchers are calling early dependence signals: 22% of students say they find it harder to work without AI, 20% say they depend on it to produce better work, and 19% say they’re retaining less information because they lean on it. Those three numbers, sitting right next to the “more challenging work” stat, are the real story of 2026 — growth and risk arriving in the same dataset.

On the teacher side, the Gallup–Walton data breaks usage into nine categories of work tasks. Teachers reported time savings ranging from 60% to 84% depending on the task, with lesson planning, drafting parent communications, and generating classroom materials topping the list. Very few teachers — 7% or fewer — said AI actually made a task take longer. Beyond time, 61% of teachers said AI gives them better insight into student learning and achievement data, and 57% said it improved the quality of their grading and feedback.

Browse Our Free AI Tools Hub for Students & Teachers

Which AI Tools Students Are Actually Choosing

Adoption numbers tell you how many people use “AI.” They don’t tell you which tool is actually sitting open in a browser tab during study hall. In Digital Education Council survey work covering more than 27,000 students, ChatGPT remains the dominant choice by a wide margin, ahead of general AI-powered search engines, standalone chatbots (used by 51% of students at some point), image generators (34%), and video generators (22%). One in five students uses an AI search engine daily; only 11% say the same about chatbots specifically, which suggests a lot of “AI use” is quieter and more utilitarian than the cheating-panic headlines imply.

On the K-12 side, purpose-built education tools are growing from a much smaller base but climbing fast. Khan Academy’s tutoring assistant Khanmigo is the most frequently cited example: multiple industry trackers, including reporting that draws on Microsoft and World Economic Forum figures, put its user base at roughly 68,000 during the 2023-24 school year, climbing to more than 1.4 million by mid-2025. That kind of curve — a purpose-built tutor scaling twenty-fold in about eighteen months — is a much better indicator of where district-approved AI is headed than general chatbot adoption numbers, since Khanmigo access typically has to clear a school or district approval process before students can log in at all.

Younger Students Are Catching Up Fast

College campuses get most of the AI-in-education headlines, but the fastest percentage-point growth in 2026 is happening in middle and high schools. Pew Research’s most recent teen survey found that the number of U.S. teens aged 13 to 17 using ChatGPT for schoolwork doubled year-over-year, from 13% in 2023 to 26% in 2024, and general awareness of the tool climbed from 67% to 79% of teens over the same window.

Age matters a lot here. Pew found 31% of 11th and 12th graders use ChatGPT for schoolwork, compared to 25% of 9th and 10th graders and 20% of 7th and 8th graders — a fairly clean climb by grade level. Household income also shapes the picture: teens in homes earning $75,000 or more were considerably more likely to use ChatGPT than teens in lower-income households, though the gap in basic awareness has been closing quickly, with lower-income teens showing the sharpest year-over-year increase in simply knowing the tool exists.

Teens themselves draw a clear line around what counts as acceptable use. In the same Pew survey, 54% said using ChatGPT to research a new topic was fine, but that support collapsed to 29% for solving math problems and just 18% for writing essays outright. In other words, the students using these tools most heavily are also the ones most aware of where the ethical line sits — even if not everyone stays on the right side of it.

The Global Split: Why the U.S. and Canada Are the Outlier

This is the part of the 2026 data that surprised me most. Everywhere outside North America, faculty intent to keep using AI in teaching is climbing hard: 92% in APAC, 89% in EMEA, and 94% in Latin America, according to the Digital Education Council. In the U.S. and Canada, that same intent figure actually fell nine points in a single year, from 76% down to 67% — the lowest of any region measured.

The sentiment gap widens further when you look at risk perception. In the U.S. and Canada, 55% of faculty said AI poses a serious risk to human intellectual development, compared to just 29% in APAC. On the student side, 66% worry AI could make learning too shallow globally — but that number jumps to 81% among U.S. and Canadian students specifically, and 43% of that group said they’d support an institution-wide AI ban. Trust between classmates is fraying too: 60% of students worldwide worry peers might misuse AI for unfair advantage, rising to 73% in North America.

Assessment design is the pressure point behind most of this. Only 28% of students globally feel their assessments reflect the skills and judgment they’ll actually need in an AI-enabled workplace, and 24% say they aren’t allowed, encouraged, or required to use AI in any assessment at all — a figure that climbs to 38% in the U.S. and Canada. Meanwhile, just 31% of faculty globally feel their institution meaningfully involves them in shaping AI policy, even though 64% say they’ve personally completed AI literacy training. The gap isn’t a lack of effort. It’s a lack of coordination between what teachers are trained to do and what institutions actually let them do in the classroom.

My Research Process

Statistics on a page are easy to misread if you’ve never sat where the students and teachers in these surveys are sitting, so I don’t just aggregate numbers for pieces like this — I go hands-on with the actual tools driving the adoption curve. That means creating real accounts on the free and trial tiers of the platforms students and teachers mention most often in these surveys — ChatGPT, Khan Academy’s Khanmigo, and Google’s NotebookLM among them — and using them the way a student or overworked teacher actually would, not the way a vendor’s demo video shows.

A few things stood out. Signing up for Khanmigo’s teacher-facing tools still requires a verifiable school email or district approval before you get full classroom features — a deliberate friction point that keeps the tool out of the hands of students posing as staff, but one that also slows down individual teachers trying to test it on their own initiative outside official district rollout. On the student research side, NotebookLM’s source-grounded approach means it won’t answer questions outside the documents you’ve uploaded, which is a genuinely different experience from an open-ended chatbot and explains why so many DEC survey respondents distinguish between “research support” and “essay writing” when they talk about acceptable AI use — the tools themselves nudge people toward different behaviors.

I take notes throughout testing and cross-reference every usage claim against the original survey or study rather than a secondary aggregator, which is why you’ll notice every statistic in this piece is attributed to a named institution — the Digital Education Council, Gallup and the Walton Family Foundation, Pew Research Center, or HEPI — rather than a vague “studies show.” This review reflects my own testing and reading of the primary research, not professional, educational-policy, or legal advice. If you’re making a purchasing or policy decision for a school or district, verify directly with the vendor and the cited research organizations.

Own an EdTech or AI Tool? Get a Human-Written Review

What This Actually Means for Schools, Parents, and Policy

Put the higher-ed, K-12, and teen data side by side and a pattern emerges: usage is outrunning guidance at every level. Students and teachers have both made up their minds that AI belongs in their workflow — the DEC researchers put it bluntly: adoption is now widespread, but coherent practice is not. The institutions still deciding whether to have a policy are already several steps behind the people they’re trying to govern.

For schools and districts, the practical takeaway from this year’s numbers isn’t “ban it” or “adopt it blindly” — it’s that the institutions seeing the least friction are the ones that paired adoption with explicit assessment redesign, not just a usage policy PDF nobody reads. Only 16% of students in some surveys say they know when AI use is appropriate because their school actually published clear guidelines — everyone else is guessing, which is exactly how “unauthorized use” numbers stay stubbornly high even in schools that technically have an AI policy on paper.

For teachers specifically, the Gallup-Walton data is genuinely encouraging if the “AI dividend” framing holds up: teachers who invest the time to become regular users are getting real hours back, and the majority of them say the quality of their work improved, not just the speed. The barrier for the other 40% who aren’t using it at all is rarely willingness — it’s training time, tool sprawl, and unclear district approval processes, three things a single well-run onboarding session can fix faster than most people expect.

For parents, the most useful number in this whole dataset might be the acceptability split from Pew’s teen survey: research support gets broad buy-in from teens themselves (54%), while essay writing does not (18%). That’s a workable conversation starter at home — not “are you using AI,” but “what are you using it for,” since the data shows most teens already have an internal sense of where the line sits even without a formal school policy telling them.

FAQ

What percentage of students use AI in 2026?

Globally, 88% of higher-education students now use AI in their learning, according to the Digital Education Council’s 2026 survey of over 45,000 respondents across 35 countries. Among U.S. teens aged 13-17, ChatGPT use specifically for schoolwork sits at 26%, per Pew Research.

How many teachers use AI tools?

In U.S. K-12 schools, 60% of public school teachers use AI tools in some capacity, and 32% use them at least weekly, according to the Gallup-Walton Family Foundation study. In global higher education, 77% of faculty report using AI in their teaching.

How much time does AI actually save teachers?

Teachers who use AI at least weekly save an average of 5.9 hours per week — roughly six weeks over a full school year — mostly on lesson planning, grading support, and drafting parent communications, per the Gallup-Walton Family Foundation survey of 2,232 U.S. public school teachers.

Is AI use in schools mostly for cheating?

No — the data suggests otherwise. Teens themselves rate research support (54% approval) far more acceptable than essay writing (18% approval), and 61% of higher-ed students say AI frees them to focus on deeper thinking rather than replacing it. The bigger problem schools report isn’t malicious cheating so much as unclear guidance on where the line sits.

Why is U.S. faculty enthusiasm for AI falling while the rest of the world’s is rising?

The Digital Education Council’s 2026 survey found faculty intent to keep using AI in teaching fell from 76% to 67% in the U.S. and Canada over one year, while it rose to 89-94% across EMEA, APAC, and Latin America. U.S. and Canadian faculty also report far higher concern about AI’s risk to intellectual development (55% vs. 29% in APAC), which researchers link to weaker institutional AI policy involvement in North America specifically.

Which AI tool do students use most for schoolwork?

ChatGPT remains the most widely used general AI tool among students by a clear margin, based on Digital Education Council survey data covering more than 27,000 students. Purpose-built classroom tools like Khan Academy’s Khanmigo are smaller in absolute numbers but have grown far faster on a percentage basis, expanding from roughly 68,000 users in the 2023-24 school year to over 1.4 million by mid-2025.

Explore more of what’s changing in AI-powered education and productivity: our breakdown of the best NotebookLM alternatives for research and study, a look at whether ChatGPT actually holds up for technical coursework, our side-by-side on AI content writing versus human writing, a practical guide to researching and writing with generative AI tools responsibly, whether Claude AI is safe and trustworthy for daily use, and our diary of what happened when we used ChatGPT to write for 30 straight days.

Bottom Line

The 2026 numbers settle a debate that’s been running since 2023: this isn’t a passing trend. With 88% of students and 77% of faculty using AI globally, and U.S. teachers reclaiming nearly six weeks a year through it, AI has moved from novelty to infrastructure in education. But infrastructure without maintenance breaks down — and right now, the maintenance (clear policy, redesigned assessments, faculty buy-in on governance) is lagging noticeably behind the adoption curve, especially in the U.S. and Canada.

If you’re a student, the data says the safest lane is using AI for research, brainstorming, and clarifying concepts — the uses your peers and teachers already broadly accept — while being deliberate about not letting it replace the writing and problem-solving reps that actually build skill. If you’re an educator or administrator, the single highest-leverage move isn’t picking the “best” AI tool; it’s closing the guidance gap so fewer than 16% of your students are left guessing what’s actually allowed.