England Teachers See AI Turning Homework Into a Thinking Test

The National Education Union's 2026 AI survey gives English schools a warning that goes beyond a simple cheating panic. In a survey of 9,408 teacher members in state schools, two-thirds of secondary teachers said pupils' critical thinking had declined because of AI use. Teachers also raised concerns about creativity, writing, spelling, conversation and independent problem-solving.

The finding does not prove that every pupil using AI is learning less. It does show that teachers are seeing enough classroom evidence to demand rules, training and better assessment design. The risk is not only that a pupil submits work they did not write. The deeper risk is that the tool removes the struggle that normally builds the skill.

The Problem Is Cognitive Offloading

Learning often depends on effort that feels inefficient in the moment: planning a paragraph, choosing evidence, sitting with confusion, rewriting a weak sentence, checking a claim, remembering a method and explaining a mistake. AI can support those steps if it is used carefully. It can also replace them before the pupil has learned how to do them.

Fluent output can therefore mislead. A polished paragraph may look like progress while the student has gained little control over grammar, argument or structure. Teachers notice the gap when pupils cannot explain, defend or reproduce the submitted work without the tool.

Teachers Are Using AI Too

The NEU data is not a story of teachers rejecting technology from the outside. Three-quarters of respondents said they now use AI tools for day-to-day work, up sharply from the previous year. Most use is practical: resource creation, lesson planning and administration. Marking remains much less common.

The difference between support and substitution matters. Teachers can see the workload value of AI while still worrying about pupil over-reliance. The argument is not technology versus tradition. It is about when the tool supports professional judgment and when it quietly substitutes for the student's own thinking.

The Policy Gap Is Too Large

The survey found that nearly half of schools had no policy for AI use by staff or students, and about two-thirds had no specific student policy. The policy gap leaves individual teachers improvising around a technology that can change homework, assessment, revision and classroom trust overnight.

Improvisation is not enough. Schools need age-appropriate rules that explain when AI is banned, when it must be declared, when it can be used for feedback and when it can support research or revision. Without clear rules, honest pupils are left guessing and dishonest use becomes harder to distinguish from permitted support.

AI Tutors Need More Than Access Claims

The government's interest in AI tutors for disadvantaged pupils sits directly inside this dispute. In principle, extra support outside normal classroom hours could help pupils who lack private tutoring or quiet study space. The NEU response shows that access alone cannot justify the idea.

Teachers worry that pupils who most need human encouragement, explanation and relationship-based support could be given a cheaper substitute. An AI tutor may help with practice questions or revision prompts. It cannot notice the same emotional, social and safeguarding signals a trained adult can see. The design has to start from that limit.

Assessment Has to Move Closer to the Process

The classroom response should not rely only on detection software. AI detectors are imperfect and can create false accusations. Better assessment design asks students to show the process: handwritten planning, oral explanation, in-class drafting, source logs, revision notes and marks for how a piece of work developed.

Those methods are not nostalgic. They are evidence. If a pupil can explain choices, defend reasoning and improve work after feedback, the teacher can see learning. If the final answer is the only artefact, AI makes that evidence thinner.

The Goal Is Better Friction

Schools should not pretend AI can be kept outside the classroom forever. Pupils will need to learn how to question outputs, check sources, spot bias, protect privacy and use AI without surrendering judgment. Such judgment is part of modern literacy.

But education cannot be redesigned around speed alone. A child who produces a clean answer in seconds may have skipped the very practice that makes future independence possible. The task for schools is to keep the useful friction: enough effort to build memory, argument, writing and patience, with AI introduced only where it strengthens those habits instead of replacing them.