Teachers Are Using AI Every Week Now. Here's How to Do It Without Losing What Matters

What AI is genuinely good at

Let's be practical. After a couple of years of teachers experimenting, a fairly clear picture has emerged of where these tools earn their place.

First drafts of routine material. A set of ten practice questions on fractions, three versions of the same reading passage at different levels, a rubric for a persuasive essay, a parent email about a trip. None of these need to be original works of art. A decent first draft in thirty seconds, which you then fix, is a real saving.

Differentiation. Rewriting a text for a weaker reader, or adding an extension task for the student who finishes in ten minutes, used to take an evening. It now takes a few minutes, as long as you check the result.

Explaining the same idea several ways. When a class doesn't get an explanation, it helps to have four more ready: an analogy, a worked example, a diagram description, a real-world case. AI is good at producing options for you to choose from.

Admin. Turning rough notes into a report comment, summarising a long policy document, drafting a newsletter. Nobody became a teacher to do these things.

What it's still bad at

The list of weaknesses matters just as much, because this is where teachers get caught out.

Facts and sources. AI tools still invent things: dates, quotations, references that sound real and aren't. Anything factual that goes in front of students needs checking against a source you trust.

Knowing your students. The tool doesn't know that Ayesha understands the maths but panics when a question is wordy, or that half your class missed last week's lesson. It produces material for an imaginary average student who doesn't exist in your room.

Judging quality of thinking. AI can mark spelling and structure. It's much weaker at noticing the student who has an unusual but correct idea, or the one whose neat answer hides a misunderstanding. That judgement is the core of teaching.

Pitching the level. Ask for "Year 8 level" and you'll often get something too easy, too hard, or oddly mixed. You still have to read it as a teacher.

A simple workflow that keeps you in charge

The teachers who use AI well tend to follow the same rough pattern, even if they've never written it down.

1. Decide the learning goal yourself

Before opening any tool, write one sentence: what should students be able to do by the end of this lesson that they can't do now? If you skip this, AI will happily fill the gap with generic content, and the lesson will drift.

2. Give the tool real context

The more specific the request, the more useful the answer. Compare "make a worksheet on photosynthesis" with "make eight questions for 13-year-olds who know that plants need sunlight but confuse photosynthesis with respiration; start easy and end with one question that asks them to explain the difference in their own words".

Never paste in students' names, grades or personal details. Describe the class, not the children.

3. Treat the output as a draft

Read every word before it reaches a student. Fix the level, cut the fluff, check every fact. A useful habit is to change at least one thing in every AI draft. If you can't find anything to improve, you probably didn't read it closely enough.

4. Put the saved time somewhere visible

This is the step most people skip, and it's the one that matters. If AI saves you forty minutes on Sunday's planning, decide in advance where that time goes: feedback conversations with three struggling students, a phone call to a parent, actually finishing on time for once. Otherwise it quietly disappears into more admin, which is exactly what the survey numbers suggest is happening.

Four requests worth saving

Most of the value comes from a handful of requests you'll use again and again. Here are four that teachers tend to keep. Adjust the details for your subject and age group.

For differentiation: "Rewrite this passage for a student reading about two years below their age. Keep every key idea, shorten the sentences, and explain any technical word in brackets the first time it appears."

For checking understanding: "Write five short questions that would show whether a student truly understands [topic], including one question that a student with the most common misconception would get wrong. Tell me which question that is and why."

For feedback: "Here is my marking rubric and an anonymised student paragraph. Suggest two specific strengths and one next step, written to the student in a warm, plain tone, under 60 words." Then rewrite it in your own voice before it goes anywhere near the student.

For planning: "I have 45 minutes to teach [goal] to [age] students who already know [prior knowledge]. Suggest three different lesson openings that take no more than five minutes each."

Notice what these have in common. Each one gives the tool a clear job, a real constraint and a reason. Each one also keeps a decision for you: which version to use, which question matters, how the feedback should sound.

The academic honesty problem

Teachers aren't only using AI. They're also dealing with students using it. In the same Bett survey, 30% of teachers said they'd suspected unauthorised AI use in their classes within just four weeks. And in England, a National Education Union survey of 9,408 teachers found two in three schools still had no student-specific AI policy.

Detection tools aren't a reliable answer; they produce false accusations, and students quickly learn to work around them. The approaches that hold up better are about designing work where AI can't do the thinking for the student:

• Ask for the process, not just the product. Plans, drafts, notes and a short reflection on how the work developed.
• Use more in-class writing and talking. A five-minute conversation about an essay shows very quickly whether the student understands it.
• Make tasks personal or local. "Interview a grandparent about how school was different" is hard to outsource.
• Teach responsible use openly. Show students how you use AI, where it failed you, and what you check. Students copy what teachers model.

A word on wellbeing

The appeal of AI for teachers is mostly about time and tiredness, and those are real problems. But a tool that makes it easier to produce more material can also make it easier to work longer. If you notice you're now making three versions of every worksheet because you can, rather than because your students need them, it's worth stepping back.

The best use of AI in teaching is often boring: it does the paperwork so you can leave school at a reasonable time and come back with energy for the part only you can do.

What students still need from you

Every survey in the last two years points in the same direction. Students are using AI heavily, and many are less excited about it than they were. A Gallup and Walton Family Foundation survey found that Gen Z's excitement about AI fell 14 points in a year. What young people seem to want isn't more machine-made content. It's help that fits them.

That's still a human skill: noticing confusion before it's spoken, knowing when to push and when to encourage, building the kind of trust that makes a student willing to try something hard in front of you. No tool does that. If anything, it's becoming the part of teaching that's most valuable, because it's the part that can't be automated.

Where this leaves teachers

AI isn't going away from classrooms, and pretending it will isn't a plan. The teachers who'll find these tools most useful are the ones who are clear about the job: use AI for drafts, admin and options; keep the goals, the judgement and the relationships for yourself.

And for teachers thinking about teaching beyond their own school, the same skills matter even more online, where a good one-to-one teacher is often exactly what a student can't get from a chatbot. If that interests you, Next Gen Remote Learning welcomes teachers who want to teach students one-to-one, from anywhere.

Frequently asked questions