76% of Teachers Now Use AI. Half Their Schools Have No Policy
A new NEU survey of over 10,000 teachers shows AI use has jumped from 53% to 76% in a year, but almost half of schools still have no policy governing it.
The numbers moved faster than the guidance
The National Education Union's latest State of Education survey, published this year and based on responses from 10,311 teachers and 2,996 support staff in English state schools, found that 76 per cent of teachers are now using AI tools for day-to-day work. A year ago, that figure was 53 per cent. Over the same period, the share of schools with any formal AI policy — for staff or for students — has barely moved. According to the NEU, 49 per cent of teachers say their school has no policy at all.
That gap is the story. Adoption has become the default while governance has stayed optional, and the two numbers are now badly out of step.
What teachers are actually doing with it
The survey, fielded between 5 and 16 February this year, breaks down where AI has embedded itself in daily practice. Resource creation leads at 61 per cent of respondents, followed by lesson planning at 41 per cent and administrative tasks at 38 per cent. This tracks with what NeuralClass has heard from schools directly: AI has become a workload tool first, used quietly by individual teachers to draft worksheets and plans, rather than something rolled out through a whole-school system.
That pattern matters for the policy question. When adoption happens bottom-up, tool by tool, teacher by teacher, it is much harder for a single school policy to catch up with what is actually happening in practice. Senior leaders may not know which tools are in use, what data is being entered into them, or how consistently marking and feedback are being AI-assisted across departments.
The NEU also found a split by phase: secondary schools are somewhat more likely to have a policy in place than primaries, which lag furthest behind. That is the inverse of what safeguarding logic would suggest, since primary pupils are the age group least equipped to evaluate AI output critically.
Teachers are also the sceptics
The same survey punctures a common assumption: that teachers using AI heavily are also its biggest cheerleaders for classroom deployment. Asked about the government's plan to expand AI tutoring, only 14 per cent of teachers expressed agreement, with just 4 per cent strongly agreeing. Nearly half — 49 per cent — said they opposed the plan outright.
There is also a live concern about what AI is doing to pupils, not just teacher workload. Two-thirds of secondary teachers (66 per cent) told the NEU that pupils' critical thinking has declined as a result of AI use, compared with 28 per cent of primary teachers — a gap plausibly explained by the fact that secondary pupils have far more independent access to chatbots for homework and revision.
Daniel Kebede, the NEU's general secretary, put it plainly: "Students must be able to think for themselves. This is at the heart of learning, but our survey shows a reliance on AI is having an effect on students' ability to think critically. AI will only improve learning and support teachers in their role if implemented correctly, within a vision of a highly skilled profession."
That sentence is worth sitting with. Teachers are not rejecting AI — three in four now use it themselves — but they are drawing a sharp line between AI as a workload tool for staff and AI as a substitute for pupil thinking. The government's tutoring ambitions sit on the wrong side of that line for most of the profession.
Why this is a governance failure, not a technology one
None of this is really about whether AI tools work. It is about the absence of the structures that would let schools use them responsibly at scale. A school where half of staff are quietly using generative AI to mark, plan and draft, with no policy on data handling, no agreed line on what pupils may use unsupervised, and no shared understanding across departments, is not "piloting" AI. It has already adopted it, informally, without the safeguards that formal adoption would require.
This is consistent with the wider pattern seen across England this year: individual enthusiasm running ahead of institutional readiness, exactly the gap that Ofsted-adjacent commentary and the Department for Education's own advisers have flagged repeatedly. The NEU survey simply puts a hard number on it: not a minority of outlier schools, but roughly half of the entire state sector.
What to do
School leaders should treat the absence of an AI policy as an active risk rather than a neutral gap, given that staff use is already widespread regardless. A minimum policy needs three things: what data may and may not be entered into AI tools, what pupils are permitted to use unsupervised versus under supervision, and who in the school is accountable for reviewing new tools before they spread informally. Primary schools in particular should not assume that lower reported usage means lower urgency; the NEU's own figures suggest primaries are furthest behind on policy, not on need.
What to watch
Watch how the Department for Education responds to the tutoring-plan scepticism revealed in this survey. A rollout that 49 per cent of the profession actively opposes will face implementation friction regardless of the evidence base behind it. Also watch next year's NEU survey for whether the policy gap closes at all — if usage keeps climbing while policy coverage stays flat, the story next year will be less about adoption and more about the incidents that an ungoverned rollout eventually produces.