Teach17

Artificial intelligence is no longer something schools can treat as a distant possibility. It is already being used by teachers to plan lessons, create resources and reduce administration. It is also being used by pupils to ask questions, complete homework and improve their writing.

Half of the teachers who responded to a Department for Education survey said they were already using generative AI. Meanwhile, research from the National Literacy Trust found that 66.5% of young people aged 13 to 18 used generative AI in 2025, most commonly to ask questions and receive help with homework.

However, change does not automatically mean improvement. The long-term evidence on AI and pupil learning is still developing. AI can help pupils complete a task successfully without necessarily helping them understand or remember the knowledge behind it.

Here are 17 ways AI is likely to change teaching.

1. AI will become a teaching assistant rather than a replacement teacher

The most realistic role for AI is not to replace teachers but to assist them.

AI can produce an initial lesson outline, suggest questions, adapt a text or generate ideas for an activity. The teacher must still decide whether the material is accurate, appropriate and worth using.

A system may be able to explain photosynthesis in five different ways, but it cannot see the puzzled expression on a pupil’s face, remember what the class struggled with last week or judge whether the room needs another example or a complete change of direction.

The OECD describes the most promising role of AI as augmenting human teaching while preserving the teacher’s professional judgement and control.

2. Lesson planning will become faster

AI can remove some of the time spent staring at an empty planning document.

A teacher could ask it to suggest a sequence of explanations, examples, retrieval questions and independent tasks. The result should not simply be copied into a lesson, but it can provide a useful first draft.

An Education Endowment Foundation trial involving KS3 science teachers found that teachers using ChatGPT for lesson preparation spent an average of 56.2 minutes a week on the task, compared with 81.5 minutes among teachers who were not using it. This represented a time saving of approximately 31%.

The greatest benefit may not be producing more resources. It may be giving teachers more time to think carefully about how they will explain difficult ideas.

3. Teachers will be able to adapt resources more quickly

Most teachers already differentiate and adapt lessons, but doing this well can take considerable time.

AI can help produce:

A simplified version of a difficult text
A vocabulary list for pupils with weaker literacy
Sentence starters for an extended answer
A more demanding version for pupils who need stretching
A translated explanation for a newly arrived pupil
A set of worked examples with gradually reduced support

This could make adaptive teaching more manageable. However, the teacher will still need to check that the adaptation has not removed essential vocabulary, introduced a misconception or made the task so simple that meaningful learning disappears.

4. Teachers will have more ways to explain difficult ideas

When a pupil does not understand an explanation, simply repeating the same words more slowly rarely solves the problem.

AI can suggest analogies, examples, diagrams, stories and alternative explanations. A science teacher could ask for three ways to explain electric current, or a history teacher could request an analogy that helps pupils understand the balance of power.

The teacher’s subject knowledge remains essential. AI-generated analogies can be memorable but scientifically or historically misleading. Teachers will need to select, improve or reject them.

Used carefully, AI may give teachers a larger collection of possible explanations without taking away the need to understand the subject deeply.

5. Classroom questioning will become more sophisticated

AI can quickly generate questions at different levels of difficulty. More importantly, it can help teachers anticipate the wrong answers pupils are likely to give.

A teacher could ask AI to produce:

Five common misconceptions about forces
A hinge question on simultaneous equations
Three plausible but incorrect explanations of evaporation
Questions that distinguish memorisation from genuine understanding
Follow-up questions for a pupil who gives a partially correct answer

This could make questioning more diagnostic. Instead of merely asking whether pupils know an answer, teachers can explore why they think it is correct.

AI-generated questions will still need checking. They can be repetitive, poorly pitched or based on incorrect assumptions about the curriculum.

6. Pupils will have access to more individual tutoring

A classroom teacher cannot provide continuous one-to-one support to 30 pupils at once. AI tutors may offer pupils immediate explanations, hints and practice questions while the teacher works with somebody else.

A well-designed tutor could avoid revealing the answer immediately. It might ask a question, provide a clue, identify an error or direct the pupil back to an earlier step.

This is where the design of the technology matters. The OECD warns that general AI tools may improve the quality of a pupil’s finished work without producing lasting learning. Educational tools are more promising when they are designed to question, prompt and guide rather than simply complete the task.

The best AI tutor should help a pupil think. It should not do the thinking for them.

7. Feedback could become quicker

AI may help teachers identify patterns in pupils’ work and prepare an initial set of feedback points.

For example, it could help sort errors into categories, identify missing vocabulary or suggest a follow-up question. It might also give pupils immediate feedback on low-stakes practice while the teacher concentrates on more complex work.

However, feedback is not valuable merely because it arrives quickly. It must be accurate, understandable and something the pupil can act upon.

Teachers should also be cautious about uploading pupils’ work. Names, personal details, assessment information and other sensitive data should not be entered into public AI tools. Department for Education guidance states that personal data should not be used in generative AI systems unless the school has ensured that the product and process comply fully with data-protection requirements.

8. Teachers will spend less time on routine administration

AI may reduce some of the repetitive writing surrounding teaching.

It could help create an initial draft of:

A letter to parents
A revision timetable
A trip information sheet
A scheme-of-work overview
Meeting notes
A classroom newsletter
A departmental action plan
A bank of report comments

The teacher remains responsible for the final document. AI does not know the pupil, family or school context unless that information is entered, and entering it may create privacy and safeguarding problems.

The sensible approach is to use AI to produce structures and general wording, while keeping confidential information out of the system and adding the important personal detail afterwards.

9. Homework will have to change

Traditional homework can no longer be assumed to show what a pupil can do independently.

A polished essay, computer program or extended answer may have been written by the pupil, heavily edited by AI or generated almost entirely by a chatbot.

Schools are therefore likely to place greater emphasis on homework that reveals thinking. This might include:

Showing drafts and corrections
Explaining how an answer was reached
Recording a brief spoken explanation
Comparing an AI response with the pupil’s own answer
Completing part of the task in class
Responding to questions about the finished work

The Department for Education has already advised schools to review homework and unsupervised-study policies in response to the availability of generative AI.

10. Assessment will focus more closely on the learning process

AI will not make assessment impossible, but it will make some forms of assessment less trustworthy.

Teachers may need to collect evidence from different stages of a task rather than judging only the polished final product. Drafts, plans, practical work, classroom writing, oral questioning and supervised tasks may all become more important.

Some pupils may also be asked to explain or defend their work. A short conversation can quickly reveal whether a pupil understands an argument, calculation or piece of code they have submitted.

Ofqual and the Joint Council for Qualifications have made clear that submitting AI-generated coursework as a pupil’s own work is malpractice. Current guidance encourages schools to strengthen supervision, policies and conversations with pupils about authenticity.

11. AI literacy will become part of every subject

Knowing how to use AI responsibly will become a basic part of education.

Pupils will need to understand:

What generative AI can and cannot do
Why it sometimes invents information
How training data can contain bias
Why personal information should not be entered
How copyright and ownership apply
When AI use must be declared
How to question and improve an output
When using AI is inappropriate

This should not be left entirely to computing lessons. In English, pupils may examine AI-generated writing. In science, they might check an explanation for misconceptions. In history, they could investigate invented sources or biased interpretations.

UNESCO’s AI competency frameworks place human responsibility, ethics, critical judgement and safe use alongside technical understanding.

12. Fact-checking will become a more important classroom skill

AI can produce confident, fluent and completely incorrect answers.

This creates a useful teaching opportunity. Instead of pretending that pupils will not use AI, teachers can show them how to challenge it.

A lesson might ask pupils to identify:

An invented quotation
A false reference
A scientific misconception
A calculation error
A one-sided argument
A claim unsupported by evidence
A real fact placed in the wrong context

The aim should be to move pupils from passive acceptance to active evaluation. National Literacy Trust research found that one in four young people using generative AI did not interact with it in a critical or creative way.

“AI said it” cannot become a replacement for evidence.

13. Writing will be taught more as a process

AI can produce grammatically fluent writing in seconds. This means schools will need to be clearer about what they are trying to teach when they ask pupils to write.

The purpose might be to develop vocabulary, organise an argument, practise sentence construction or communicate an original idea. Allowing AI to complete the whole task may produce an impressive paragraph while removing the learning.

Used carefully, AI could still support writing. Pupils might:

Compare two possible openings
Ask for feedback on the clarity of an argument
Identify weaknesses in an AI-generated paragraph
Improve an inaccurate model answer
Compare an AI version with their own voice
Use suggestions before deciding what to accept or reject

The National Literacy Trust argues that AI should support literacy and self-expression rather than replace them. It also reports concerns among teachers about overreliance, reduced concentration and pupils taking shortcuts.

14. AI could improve accessibility for some pupils

AI has the potential to remove some barriers experienced by pupils with SEND.

Tools may help by:

Converting speech into text
Reading text aloud
Explaining vocabulary
Breaking instructions into smaller steps
Creating visual supports
Adjusting the reading level of a passage
Providing repeated practice without embarrassment
Supporting communication and organisation

This could help pupils access the same curriculum with more appropriate support.

However, accessibility should not be confused with simply reducing difficulty. Pupils still need ambitious teaching, and AI tools must be selected carefully. An OECD review of AI for pupils with special educational needs identifies genuine opportunities but also stresses the importance of evidence, ethical design, data security, monitoring and accountability.

15. AI could either reduce or widen educational inequality

AI may give pupils access to explanations and support that previously required private tuition. That could help pupils who cannot afford additional teaching.

The opposite is also possible. Some pupils will have paid tools, newer devices, reliable internet access and adults who can show them how to use AI effectively. Others may have limited access or use free tools without understanding their weaknesses.

There may also be a difference between pupils who use AI to deepen their thinking and those who use it to avoid thinking.

The Education Endowment Foundation has warned that education technology can widen disadvantage when access, implementation and teaching quality are uneven. The National Literacy Trust similarly identifies access to devices, connectivity and digital literacy as important barriers.

Schools will need to teach AI use rather than assuming every pupil will learn it independently.

16. Strong subject knowledge will become more valuable, not less

It may appear that AI makes subject expertise less important because information can be generated instantly. In reality, teachers need knowledge to judge whether an output is any good.

AI frequently produces explanations that are almost correct. These can be more dangerous than obvious nonsense because they sound convincing.

A knowledgeable teacher can notice that:

A science analogy breaks down
A historical example is anachronistic
A quotation has been invented
A mathematical method contains a hidden error
A text is pitched at the wrong level
An exam question does not match the specification

Teachers who understand their subject will be able to use AI selectively and confidently. Those who depend on it to provide the knowledge may struggle to recognise its mistakes.

17. Human relationships will become even more important

The more content technology can produce, the more valuable the genuinely human parts of teaching may become.

Teaching involves noticing when a pupil is unusually quiet, restoring confidence after failure, managing behaviour, creating a sense of belonging and persuading a reluctant class that a difficult idea is worth understanding.

AI cannot take responsibility for safeguarding a child. It cannot build the same shared history with a class, make a professional ethical judgement or understand all the small details behind a pupil’s behaviour.

The real promise of AI is not a classroom with fewer teachers. It is a classroom where teachers spend less time producing routine documents and more time explaining, questioning, listening and working directly with pupils.

AI will change many of the tools used in teaching. It should not change the fact that learning remains a deeply human process.