AI for teachers: a practical guide to planning, feedback and academic honesty

AI for teachers means using artificial intelligence tools, mostly generative AI chatbots and the AI features now built into everyday software, to take on the drafting and adapting work that surrounds teaching: first drafts of lesson resources, quiz questions, differentiated texts, feedback comments and messages to parents. Used well, it gives you a faster starting point. It does not replace your judgement about what pupils need, and everything it produces needs checking before it reaches a classroom.

This guide covers what AI can and cannot do for a teacher, how to use it for planning, feedback and differentiation, how to handle AI in pupils’ work, and the risks worth taking seriously. The examples are written with multilingual international schools in mind, where one class may include pupils working in three or four home languages and families who expect regular, clear communication.

What is AI for teachers?

Most of what teachers call “AI” today is generative AI: software trained on very large amounts of text, and sometimes images, that produces new content in response to an instruction called a prompt. Teachers meet it in three forms:

• General-purpose chatbots that answer questions and draft text on almost any topic.
• AI features inside tools you already use, such as presentation software, learning platforms, word processors and translation tools.
• Teacher-specific platforms that package prompts into ready-made tools for rubrics, lesson outlines or report comments.

All three work in a similar way underneath. They generate plausible text; they do not verify facts, know your pupils or understand your curriculum unless you tell them. That single point explains most of the advice below. AI is useful for producing drafts quickly and unreliable at deciding what is true or appropriate, so the teacher’s job shifts from writing everything from scratch to specifying clearly and checking carefully.

What the evidence says so far

Much research on AI in education is small-scale, so one of the more rigorous studies is worth knowing. In a randomised trial run by NFER and commissioned by the Education Endowment Foundation, 259 Year 7 and Year 8 science teachers from 68 schools took part. Teachers who used a generative AI chatbot, together with a short guide on using it, spent around 56 minutes a week on lesson and resource preparation, compared with 81.5 minutes for teachers who did not use AI: an average saving of 25.3 minutes a week. An expert panel that did not know which group had produced which resources found no evidence that their quality differed.

Two details matter. Teachers mostly used AI for one or two activities rather than whole lessons, most often to create questions and quizzes or to find ideas for activities. And the trial measured preparation time and resource quality, not pupil learning. A fair reading is that AI can save some planning time on specific tasks without lowering quality, provided a teacher checks the output.

How to use AI for lesson planning

What you get back depends on the context you give. A useful prompt usually has four parts:

1. Context: year group, curriculum, what pupils already know, and anything about the class that affects the lesson, such as the range of English proficiency.
2. Task: exactly what you want produced, for example ten retrieval questions, three worked examples or a starter activity.
3. Constraints: length, reading level, vocabulary to include or avoid, and time available.
4. Format: a table, a numbered list, or questions with answers underneath.

Worked example. The prompt below is an illustration of how a Year 8 science teacher in a British-curriculum school might ask for help with a lesson on the particle model:

“I teach Year 8 science (ages 12 to 13) in an international school. Pupils have studied states of matter but many think particles in a solid do not move. About a third of the class are still developing academic English. Write eight multiple-choice questions on the particle model of solids, liquids and gases. Each question should have one correct answer and three wrong answers based on common misconceptions. Use short sentences and avoid idioms. At the end, add a table showing the answers and the misconception each wrong option tests.”

The result is a draft. Before using it, check each question against your own subject knowledge, make sure the wrong options reflect misconceptions your pupils actually hold, and remove anything that relies on cultural knowledge the class may not share. Then ask for revisions in plain language: “Make questions 3 and 6 harder” or “Replace the example about snow with one pupils in Dubai would see every day.”

AI is also useful earlier in planning as a thinking partner. You might ask it to list misconceptions pupils commonly hold about a topic, suggest three ways to sequence a unit, or critique your own lesson plan against its success criteria. Treat its suggestions as options to weigh, not answers to adopt.

Using AI for feedback without losing your voice

Feedback is where AI can save a great deal of time and where it needs the most care, because feedback only helps when it is accurate, specific and linked to what the pupil should do next. Uses that tend to work well:

• Comment banks built from your own criteria. Give the tool your success criteria and ask for short, specific comments for each level of performance, then edit them into your own words.
• Whole-class feedback. Read the work yourself, note the three or four most common strengths and errors, and ask AI to turn your notes into a one-page feedback sheet with a short practice task for each error.
• Practice tasks for common mistakes. If many pupils misused the past perfect tense, ask for ten sentences that target that exact error.
• Model answers at different standards. Ask for a weak, a secure and a strong answer to the same question so pupils can compare them against the criteria.

Two rules keep this safe. First, do not paste pupils’ names, personal details or identifiable work into a tool your school has not approved for that purpose. Second, the judgement stays with you: decide the mark or level yourself, and use AI only to help express feedback you have already decided on.

Differentiation and multilingual classrooms

In international schools, differentiation often means making the same content accessible to pupils at very different stages of English. AI handles several of these tasks quickly:

• Adapting reading level. Ask for the same text at two or three reading levels, keeping the key subject vocabulary unchanged, so every pupil studies the same content.
• Glossaries and sentence frames. Generate a key-word list with simple definitions and example sentences, or sentence starters for explaining, comparing and justifying.
• Home-language support. Produce a bilingual glossary in, for example, Arabic, Hindi or Tagalog. Subject-specific terms can be mistranslated, so ask a colleague or a pupil who speaks the language to check it before you print it.
• Communication with families. Draft clear, jargon-free messages and translated versions for families who prefer another language, then check the tone as well as the meaning. A message that reads as neutral in English can sound abrupt in translation.

Differentiation should adjust the support, not the ambition. If an AI-simplified text drops the concept you are teaching, it has simplified too far.

How to detect AI in pupils’ work, and why task design matters more

Teachers often ask how to detect AI, and the honest answer is that no single method settles it. Some schools use detection software; this guide makes no judgement about how accurate any particular tool is. Whatever your school uses, treat a flag as a reason to look more closely and talk to the pupil, never as proof on its own, and follow your school’s academic honesty policy.

Signs that justify a conversation include:

• writing that differs sharply from the pupil’s in-class work in vocabulary, structure or accuracy;
• references, quotations or figures that do not exist or cannot be traced;
• generic content that ignores the texts, examples or discussions from your lessons;
• a pupil who cannot explain or extend their own argument when asked.

The conversation is usually more informative than any single sign. Ask the pupil to talk you through how they produced the work, what they would change, or to explain one paragraph in their own words.

Over time, assessment design does more to protect academic honesty than detection. Practical options include:

• Collect baseline writing in class early in the year, so you know each pupil’s voice.
• Assess the process as well as the product: plans, drafts, annotated sources and version history.
• Tie tasks to your lessons, for example by asking pupils to use the results of the experiment the class carried out.
• Add a short oral element, such as a two-minute explanation of the submitted work.
• State the AI rules for each task. A simple scale works: “no AI”, “AI for ideas only, with acknowledgement”, or “AI allowed, with the prompts attached”.

Pupils also need to be taught how to use AI honestly: how to acknowledge it, how to check what it says, and why handing in unedited AI text defeats the purpose of the task.

Limits and risks of AI in education

• Confident errors. Generative AI can produce false facts, invented sources and flawed worked solutions in fluent prose. Check everything, especially in maths and science.
• Bias and narrow examples. Outputs can default to one country’s names, settings and assumptions. In a multilingual class this matters, so ask for varied, local examples and read for stereotypes.
• Privacy. MIT Sloan’s practical guide for instructors is direct: “You should never enter any data or input that is confidential or sensitive into publicly accessible generative AI tools.” Remove names and identifying details, and use tools your school has approved.
• Age terms and access. Many AI tools set minimum ages in their terms of use, and not every family has the same access at home. Check both before setting AI-based homework.
• Deskilling. If AI drafts every plan, you lose the thinking that planning builds. Keep some planning fully your own, especially for new topics.
• Workload creep. Faster resource production can turn into more resources rather than less work. Decide in advance what the saved time is for.

A four-week plan for getting started

If you are new to AI, a short, structured trial works better than trying everything at once. One way to organise a first month:

Frequently asked questions

What is AI for teachers?

AI for teachers refers to using artificial intelligence tools, mainly generative AI, to support the work around teaching: planning, creating and adapting resources, drafting feedback and communicating with families. The teacher sets the task, checks the output and makes the professional decisions. It works as a drafting assistant, not as a source of truth or a substitute for knowing your pupils.

Can AI replace teachers?

AI can produce explanations, questions and feedback text, but teaching involves far more: knowing individual pupils, noticing confusion, building relationships, managing a room and judging what a learner needs next. Those remain human work. A more realistic change is that AI takes on some drafting and adapting tasks, giving teachers more time for the parts of the job only they can do.

How can teachers detect AI in pupils’ work?

No single method settles it. Compare the work with the pupil’s in-class writing, check that any references exist, and look for content that ignores what was taught. Above all, ask the pupil to explain their work. Treat any detection tool’s result as a prompt for that conversation rather than proof, and follow your school’s academic honesty policy.

Is there free AI for teachers?

Many general AI tools and teacher platforms offer free versions, often with limits on use or features. Before relying on one, check how it handles what you type in, whether your school has approved it, and its minimum age terms if pupils will use it. Free or paid, the output needs the same careful checking.

If you want to study this area in depth, Taito Education offers a master’s in AI in education: the Master of Education (AI in Education), awarded by Woolf University, Malta, and studied over 12 months. The six-month Postgraduate Certificate in Education (AI in Education), also awarded by Woolf University, is available as an AI in education certificate, and the three-year doctorate in AI in education is awarded by Universidad Tecnológica Atlántico Mediterráneo (UTAMED), Spain. All three are studied fully online.