Teacher AI Use Statistics 2026
By Axis Intelligence Research
Co-author: Sarah Mitchell, AI & Machine Learning | Last updated: September 13, 2026 | License: CC BY 4.0
Sixty percent of U.S. public school teachers used an AI tool for their work in the 2024-25 school year, and 32% used one at least weekly. Only 18% have received any formal written guidance on how to use it. Axis Intelligence Research finds that 79% of measured classroom AI use in American schools sits outside any written school policy.
Quick Answer
Teacher AI adoption in the United States reached 60% in the 2024-25 school year, with 32% of teachers using AI weekly and saving a self-reported 5.9 hours per week. Governance has not followed. The Teacher AI Governance Index (TAGI) — an Axis Intelligence Research measure of how much classroom AI use is covered by formal written policy — reads 21.0 out of 100 at its baseline, meaning roughly four in five instances of teacher AI use happen with no written rule attached. England shows the same shape: 44% teacher adoption, and 49% of schools with no AI policy for staff or students.
Key Findings
- Axis Intelligence Research finds the Teacher AI Governance Index for U.S. K-12 schools reads 21.0 out of 100 as of the February-March 2026 Gallup wave, meaning 79% of measured classroom AI use is unaccompanied by formal written policy.
- According to Gallup’s June 2025 report for the Walton Family Foundation, 60% of U.S. public school teachers used an AI tool during the 2024-25 school year and 32% used one at least weekly.
- According to Gallup’s May 2026 findings, just 18% of U.S. K-12 teachers receive any formal written guidance on AI use, while 34% receive no guidance of any kind across ten measured work tasks.
- Axis Intelligence Research finds the governance gap is widest on lesson preparation, where 54% of teachers use AI but only 9% have a written policy for it — a 45-percentage-point gap, the largest of any measured task.
- According to the UK Department for Education’s Technology in Schools Survey published in November 2025, 44% of teachers in England used generative AI for school activities, with 35% using it for lesson planning and 5% for marking.
How Many Teachers Use AI in 2026?
The adoption curve is no longer in dispute. Gallup, surveying 2,232 U.S. public school teachers through the RAND American Teacher Panel between March 18 and April 11, 2025, found that six in ten teachers had used an AI tool for their work that school year. Thirty-two percent used AI at least weekly; another 28% used it monthly or less.
RAND’s own panel data, published September 30, 2025, puts the jump in sharper relief: among English language arts, math and science teachers, AI use rose from 25% to 53% in a single year. That is a doubling, not a drift.
Adoption is also strongly graded by phase. RAND’s pre-K teacher survey recorded 29% adoption among pre-kindergarten teachers against 69% among high school teachers in the same 2024-25 year — a 40-point spread inside one profession.
Table 1. Teacher AI adoption benchmarks, 2024-2026
| Population | Adoption | Period | Source |
|---|---|---|---|
| All U.S. K-12 public school teachers | 60% | 2024-25 school year | Gallup / Walton Family Foundation |
| U.S. teachers using AI weekly or more | 32% | 2024-25 school year | Gallup / Walton Family Foundation |
| U.S. ELA, math and science teachers | 53% | 2025 | RAND Corporation |
| Same population, prior wave | 25% | 2024 | RAND Corporation |
| U.S. pre-K teachers | 29% | 2024-25 school year | RAND Corporation |
| U.S. high school teachers | 69% | 2024-25 school year | RAND Corporation |
| Teachers in England | 44% | 2024-25 school year | UK Department for Education |
| NEU member teachers, England | 76% | February 2026 | National Education Union |
Sarah Mitchell: The spread across those rows is the finding, not a data problem. Gallup measures any use across nine defined work tasks; the DfE asks about generative AI for school activities; the NEU asks members which of several tasks they reach for AI on. Three instruments, three denominators, three numbers — and all of them point the same direction at roughly the same slope. The instinct to average them into one tidy headline is exactly the instinct to resist. What survives triangulation is the shape: adoption is majority-level in secondary phases, still minority-level in early years, and climbing fast enough that any single figure has a short shelf life.
What Is the Teacher AI Governance Index (TAGI)?
TAGI — the Teacher AI Governance Index — is an Axis Intelligence Research metric that answers a question adoption statistics cannot: how much of the AI use actually happening in classrooms is covered by a written rule?
It is expressed on a 0-100 scale, where 100 would mean every teacher using AI for a given task has a formal written policy governing that task, and 0 would mean none do.
The formula, with inputs disclosed:
TAGI = (Σ formal written guidance rate across tasks ÷ Σ task use rate across tasks) × 100
For the nine work tasks measured identically in both Gallup waves:
- Sum of per-task use rates: 319
- Sum of per-task formal-guidance rates: 67
- TAGI = 67 ÷ 319 × 100 = 21.0
Both inputs come from the same instrument — Gallup’s Teaching for Tomorrow series, fielded through the RAND American Teacher Panel, weighted to National Center for Education Statistics targets, with a ±2.5-point margin of error in each wave. Task wording is identical across waves. That methodological continuity is what makes the ratio meaningful rather than decorative.
The use rates come from the wave fielded March-April 2025; the guidance rates from the wave fielded February-March 2026. Because adoption has continued climbing across that eleven-month interval, 21.0 should be read as a ceiling. The true coverage figure is very likely lower.
Governance Coverage by Classroom Task
Table 2. Teacher AI Governance Index by work task, U.S. K-12
| Work task | Teachers using AI | Formal written guidance | Governance gap | TAGI score |
|---|---|---|---|---|
| Preparing to teach | 54% | 9% | 45 pts | 16.7 |
| Making worksheets, assignments, activities | 51% | 8% | 43 pts | 15.7 |
| Modifying materials for student needs | 42% | 8% | 34 pts | 19.0 |
| Making assessments, quizzes, exit tickets | 41% | 9% | 32 pts | 22.0 |
| Administrative work | 39% | 7% | 32 pts | 17.9 |
| Supplementing instruction | 27% | 7% | 20 pts | 25.9 |
| Grading and feedback | 25% | 8% | 17 pts | 32.0 |
| One-on-one tutoring | 20% | 5% | 15 pts | 25.0 |
| Analyzing student learning data | 20% | 6% | 14 pts | 30.0 |
| National baseline | — | — | 28 pts mean | 21.0 |
Use rates: Gallup, March-April 2025. Guidance rates: Gallup, February-March 2026. TAGI and gap columns calculated by Axis Intelligence Research.
The counterintuitive result is in the right-hand column. Governance coverage is highest on the tasks teachers touch least. Grading scores 32.0 and student-data analysis 30.0 — the two best-covered tasks on the board — not because districts wrote careful grading policies, but because so few teachers use AI there that the small policy base covers a larger share of it.
Split the nine tasks into the two things teachers actually do with AI and the pattern is unambiguous:
- Preparation and back-office tasks: TAGI 18.1 (227 points of use, 41 points of policy)
- Student-facing tasks: TAGI 28.3 (92 points of use, 26 points of policy)
Sarah Mitchell: Read that as a design failure rather than a paradox. Policy has attached itself to the use case administrators worried about — a machine assigning a grade — while the use case that scaled is lesson prep. Fifty-four percent of teachers are running instructional materials through a model with a 9% chance that anyone wrote down what that means. The exposure isn’t dramatic; it’s cumulative. Worksheets carry factual claims, reading levels, examples, framing. A model that fabricates a date inside a slide deck doesn’t trigger an incident report — it just enters the curriculum. The governance conversation went where the headline risk was and missed where the volume went.
Where Guidance Exists, It Mostly Says Nothing
The 18% figure understates the vacuum, because Gallup also asked what the guidance says. Across every measured task, the most common answer is that guidance neither encourages nor discourages AI use.
Table 3. Direction of AI guidance, among teachers who receive any
| Task | Encouraged | Neither | Discouraged | Source |
|---|---|---|---|---|
| Modifying materials for student needs | 58% | 40% | 2% | Gallup |
| Making worksheets and assignments | 54% | 44% | 3% | Gallup |
| Preparing to teach | 53% | 44% | 2% | Gallup |
| Analyzing student learning data | 50% | 47% | 3% | Gallup |
| Administrative work | 44% | 53% | 4% | Gallup |
| Supplementing instruction | 42% | 51% | 7% | Gallup |
| Grading and feedback | 36% | 54% | 10% | Gallup |
| One-on-one tutoring | 35% | 57% | 8% | Gallup |
| Coaching on your own teaching | 30% | 63% | 7% | Gallup |
Grading carries the highest discouragement rate on the board at 10% — still a small number, and still the only task where a meaningful minority of schools have said don’t. Forty-eight percent of teachers receive informal guidance only: a hallway conversation, a shared norm, a department-meeting aside.
How Much Time Does AI Actually Save Teachers?
Teachers who use AI weekly estimate they save 5.9 hours per week, which Gallup converts across a 37.4-week school year into the six-weeks headline that has travelled further than any other number in this field.
The task-level detail is more useful than the aggregate.
Table 4. Share of AI-using teachers reporting time saved, by task
| Task | Saves time | Source |
|---|---|---|
| Making worksheets, assignments, projects | 84% | Gallup |
| Making assessments, quizzes, exit tickets | 83% | Gallup |
| Administrative work | 81% | Gallup |
| Preparing to teach | 80% | Gallup |
| Grading and feedback | 79% | Gallup |
| Modifying materials for student needs | 74% | Gallup |
| Supplementing instruction | 65% | Gallup |
| Analyzing student learning data | 63% | Gallup |
| One-on-one tutoring | 60% | Gallup |
Quality reports run lower than time reports across the board. Seventy-four percent of AI-using teachers said administrative work improved in quality; only 57% said the same of grading and feedback — the widest quality-confidence gap between any two tasks in the series.
Sarah Mitchell: The 5.9 hours is self-reported, estimated in half-hour increments, task by task, then summed. That is not a time-and-motion study and Gallup does not claim it is. What makes it credible is the internal consistency — the tasks with the highest reported time savings are the tasks with the highest adoption, which is the ordering you would expect if teachers were reporting something real rather than reporting enthusiasm. What makes it worth hedging is that no one has yet published a logged, observed measurement against it. Until someone does, treat six weeks as the profession’s own estimate of its dividend, not as an audited figure.
Is the AI Governance Gap the Same Outside the United States?
It is, and the convergence is the most citable thing in this dataset.
The UK Department for Education’s Technology in Schools Survey for 2024-25, fielded January-May 2025 with 1,211 teachers weighted to the School Workforce Census, recorded 44% teacher adoption in England. The task breakdown is where it matters: 35% for lesson planning, 7% for delivering live lessons, 5% for marking.
That is a 7:1 ratio of preparation use to marking use. Axis Intelligence Research computes the equivalent U.S. ratio from the Gallup task table at 2.7:1 (preparing to teach, 54%, against one-on-one tutoring, 20%).
The National Education Union’s February 2026 survey of roughly 10,300 teachers in English state schools, reweighted to the School Workforce Census, put the ratio higher still: 61% using AI for resource creation against 7% for marking — 8.7:1 — with overall adoption at 76%, up from 53% a year earlier.
Table 5. Preparation-weighted AI use, England vs. United States
| Measure | England (DfE) | England (NEU) | United States (Gallup) |
|---|---|---|---|
| Overall teacher adoption | 44% | 76% | 60% |
| Heaviest use case | Lesson planning, 35% | Resource creation, 61% | Preparing to teach, 54% |
| Lightest measured use case | Marking, 5% | Marking, 7% | Tutoring / data analysis, 20% |
| Preparation-to-practice ratio | 7.0:1 | 8.7:1 | 2.7:1 |
| Schools with no AI policy | ~80% (policy in 20% primary, 34% secondary) | 49% with no staff or student policy | 82% of teachers without formal guidance |
Sources: UK Department for Education Technology in Schools Survey 2024-25; National Education Union State of Education 2026; Gallup / Walton Family Foundation. Ratios calculated by Axis Intelligence Research. Denominators differ between instruments and the ratios are not merged.
The English policy picture matches the American one closely enough to be uncomfortable. Around one fifth of English schools had a policy on safe and appropriate AI use — 20% of primaries, 34% of secondaries. School leaders were planning AI investment for teachers over pupils by 58% to 20%. Restrictions cluster on students, not staff: 73% of schools restrict student AI use in lessons, against 34% restricting teacher use in lesson planning.
Sarah Mitchell: Two national systems with different funding models, different inspection regimes and different vendor markets arrived at the same distribution of AI use inside eighteen months. That rules out a lot of explanations. It is not a procurement story — England buys through multi-academy trusts and the U.S. through districts. It is not a policy story, because neither system had much policy. The likeliest driver is the workload contour itself: preparation is the task teachers can offload without anyone else’s consent, and marking is the task where a visible error costs them credibility with a student the next morning. Teachers are doing a risk calculation their institutions never gave them.
Which Schools Are Getting AI Guidance, and Which Are Not?
Formal policy is roughly flat across school poverty levels — 10% in the wealthiest quartile, 11% in the least wealthy, for preparation tasks. The divergence is in informal guidance, which runs 52% in the wealthiest schools against 40% in the least wealthy.
The training picture is starker. RAND’s district panel found that by fall 2024, 67% of low-poverty districts had trained teachers on AI use against 39% of high-poverty districts — a 28-point gap that district leaders’ own projections expected to persist into 2025-26.
Table 6. AI support by school and district resource level
| Measure | Higher-resource | Lower-resource | Gap | Source |
|---|---|---|---|---|
| Formal guidance, preparing to teach | 10% | 11% | −1 pt | Gallup |
| Informal guidance, preparing to teach | 52% | 40% | 12 pts | Gallup |
| District provided teacher AI training | 67% | 39% | 28 pts | RAND Corporation |
The pattern is consistent: written policy is scarce everywhere, and the advantage wealthier schools hold is in the unwritten layer — the conversations, the norms, the colleague who has already worked it out.
What Do Parents, Students and Administrators Think?
RAND’s five-population triangulation surfaced the largest perception gap in the dataset: 61% of parents believe greater AI use will harm students’ critical-thinking skills, against 22% of district leaders — a 39-point divide between the people buying the tools and the people whose children use them.
Teachers in England report the decline directly. Sixty-six percent of secondary teachers in the NEU survey agreed that pupils’ critical thinking has declined because of AI use, against 28% in primary. Notably, teachers in their twenties — the heaviest AI users in the profession — agreed at 57%, against 39% among those aged 50 and over.
Other benchmarks from the RAND panels:
- 45% of principals reported school or district AI policies or guidance
- 34% of teachers reported a school or district AI policy tied to academic integrity
- 35% of district leaders reported providing students with AI training
- Over 80% of students said teachers did not explicitly teach them how to use AI for schoolwork
That last figure pairs directly with our student AI usage statistics, which tracks the demand side of the same vacuum.
Globally, the institutional layer is barely built. UNESCO published AI competency frameworks for teachers and students in 2024, noting that only 15 countries had included AI learning objectives in national curricula as of 2022. On February 24, 2026, the U.S. House Committee on Education and the Workforce held a hearing titled Building an AI-Ready America: Teaching in the AI Age — the first congressional session dedicated specifically to teacher AI practice.
Methodology
Sources. Every figure in this article comes from a document fetched and read on September 13, 2026. Primary sources: Gallup / Walton Family Foundation Teaching for Tomorrow (two waves, June 2025 and May 2026); RAND Corporation RR-A4180-1, RR-A956-31 and RR-A4412-2; UK Department for Education Technology in Schools Survey 2024-25; National Education Union State of Education 2026; UNESCO; U.S. House Committee on Education and the Workforce. No aggregator, content-marketing roundup or secondary compilation is the source of any figure here.
TAGI construction. Nine work tasks appear with identical wording in both Gallup waves. Task use rates are derived as 100 minus the “Never” share in the 2025 per-task frequency table. Formal guidance rates are taken directly from the 2026 per-task guidance table. TAGI is the exposure-weighted ratio of the two sums, which weights each task by how much AI use actually occurs there rather than treating a task used by 20% of teachers as equal to one used by 54%. The unweighted mean coverage across the nine tasks is 22.7, and the mean per-task governance gap is 28.0 percentage points; both are published alongside the weighted figure so anyone can reproduce either.
Excluded task. Gallup’s 2026 wave measures a tenth task — using AI to get feedback or coaching on one’s own teaching, 7% formal guidance — with no counterpart in the 2025 use table. It is excluded from TAGI rather than estimated.
Wave offset. Use rates are as of April 2025 and guidance rates as of March 2026. Adoption rose over that interval, so the denominator is understated and TAGI 21.0 is an upper bound.
Cross-national figures are not merged. England and U.S. instruments use different denominators, sampling frames and task definitions. They are presented side by side, and the preparation-to-practice ratios are computed within each instrument, never across them.
A figure we did not publish. A widely repeated claim that 70% of teachers do not feel prepared to use AI, and that over 90% want more leadership support, appears in the February 2026 congressional hearing statement without an attributed survey. We could not trace either figure to a primary instrument and have left both out.
About This Dataset
Contents. 152 observations covering U.S. and English teacher AI adoption by task, frequency, school phase, teacher age and school resource level; formal and informal governance coverage; reported time savings and quality effects; and the Teacher AI Governance Index with its per-task components.
Format. UTF-8 CSV, one row per observation. Every row carries source_org, source_document, source_url, retrieved_date, is_primary and axis_calculated. Rows where axis_calculated = yes carry a method_note stating the formula.
License. CC BY 4.0. Free to use, republish and build on with attribution.
Citation. Axis Intelligence Research, Teacher AI Use Statistics 2026: Adoption, Workload and the Governance Gap, 2026.
Cite this page
APA — Axis Intelligence Research. (2026). Teacher AI use statistics 2026: Adoption, workload and the governance gap. https://axis-intelligence.com/teacher-ai-use-statistics/
MLA — Axis Intelligence Research. “Teacher AI Use Statistics 2026: Adoption, Workload and the Governance Gap.” Axis Intelligence, 2026, axis-intelligence.com/teacher-ai-use-statistics/.
Chicago — Axis Intelligence Research. “Teacher AI Use Statistics 2026: Adoption, Workload and the Governance Gap.” Axis Intelligence, 2026. https://axis-intelligence.com/teacher-ai-use-statistics/.
Frequently Asked Questions
Do teachers who use AI weekly really save six weeks a year?
That figure is the profession’s own estimate, not an audited measurement. Gallup asked weekly AI users to estimate time saved or added on each of nine tasks in half-hour increments, summed the non-overlapping totals to 5.9 hours per week, and multiplied across a 37.4-week school year. The internal consistency is reassuring — the highest-adoption tasks report the highest savings — but no published study has observed and logged those hours independently. Treat six weeks as a credible self-report awaiting verification.
Which classroom AI tasks are most likely to have a written school policy?
None of them reach one in ten. Formal written guidance peaks at 9% for preparing to teach and making assessments, and bottoms out at 5% for one-on-one tutoring. Across all ten tasks Gallup measured in 2026, fewer than one teacher in ten has a written policy for any specific activity.
Is AI grading allowed in U.S. public schools?
Mostly it is neither allowed nor forbidden — it is unaddressed. Among teachers who receive any guidance on AI grading, 54% say it neither encourages nor discourages the practice, 36% say it encourages it, and 10% say it discourages it. Grading carries the highest discouragement rate of any measured task, which still leaves nine in ten receiving no prohibition. Fifty-eight percent of teachers receive no guidance on AI grading at all.
Are teachers in high-poverty schools getting less AI guidance?
Not on paper — formal written policy is roughly equal across school poverty quartiles, at 10-11% for preparation tasks. The divergence is informal: 52% of teachers in the wealthiest schools receive informal guidance on preparation tasks against 40% in the least wealthy. The larger gap is upstream in training, where RAND found 67% of low-poverty districts had trained teachers on AI against 39% of high-poverty districts.
Why do England and the United States show the same prep-heavy AI pattern?
Both systems show AI use concentrated in lesson preparation and resource creation, and thin in marking and live instruction — a 7:1 ratio in England’s DfE data, 8.7:1 in the NEU’s 2026 survey, 2.7:1 in the U.S. Gallup task table. The systems differ in funding, procurement and inspection, which rules out most institutional explanations. The common factor is the workload contour: preparation is the task a teacher can offload alone, while an error in marking is visible to a student immediately.
Does informal AI guidance count as a policy?
Not for most compliance or audit purposes. Gallup distinguishes formal guidance — written policy or official guidance — from informal guidance such as verbal conversations and shared norms. Forty-eight percent of U.S. teachers receive informal guidance only. Informal guidance is also weaker in effect: 69% of teachers with formal guidance on making student materials say it encourages AI use, against 51% of those with informal guidance.
How fast is teacher AI adoption actually growing?
Roughly a doubling per year at the last measured interval. RAND recorded ELA, math and science teacher use rising from 25% to 53% between 2024 and 2025. The NEU recorded English teacher use rising from 53% to 76% between February 2025 and February 2026, with the share using no AI tools at all falling from 47% to 24%.
What would move the TAGI score?
Only new formal policy, or a collapse in use. The index divides written-guidance coverage by actual use, so districts publishing written policies on the high-volume preparation tasks would move it fastest — a written policy on lesson preparation alone covers 54 points of use, more than twice what a grading policy covers. Adoption rising without new policy pushes the score down.
Which age groups of teachers use AI most?
Early-career teachers lead on the tasks closest to students. The DfE found teachers under 35 in England more likely to use generative AI for lesson planning (43% against 32%) and for written feedback (21% against 12%). Teachers with under three years of experience used AI for written feedback at 27% against 14% for everyone else — the single largest experience-based gap in the DfE data.
Are teachers themselves confident that AI helps students?
Their confidence tracks the task. Seventy-four percent of AI-using teachers say it improves the quality of administrative work; only 57% say the same of grading and feedback. On student outcomes the profession is more sceptical than its administrators: 66% of English secondary teachers report a decline in pupils’ critical thinking attributed to AI, and teachers in their twenties — the heaviest users — agree at a higher rate (57%) than those over 50 (39%).