Understand AI before you teach with it.
A practical introduction for teachers. Learn how generative AI produces text, why it can be confidently wrong and how the context you give changes the draft you get back. Then practise the review habit that keeps you in charge.
- 6 short modules
- ~4 hours with practice
- 6 practice tasks
By the end, you will be able to…
Five practical capabilities you can use straight away, whichever AI tool your school uses.
- 01
Explain in plain language how a large language model produces text, and why it can be confidently wrong.
- 02
Write a teaching prompt that includes learners, curriculum, a checkable objective and the format you need.
- 03
Review an AI draft for accuracy, curriculum fit and suitability, and decide what to keep, change or remove.
- 04
Protect learners by keeping identifying information out of prompts and following school policy.
- 05
Choose which preparation tasks are worth trying with AI first, and which stay with the teacher.
Where a draft comes from, and where you come in.
- Teacher
Your prompt
Learners, curriculum, objective and format.
- AI
The model predicts
It writes a likely response, word by word, from patterns in its training.
- AI
A first draft
Fluent and organised, but unchecked.
- Teacher
Your review
Keep, change or remove. Check facts, fit and suitability.
- Teacher
Your classroom
A resource you have made your own.
AI helps at one stage of the process. The context at the start and the review at the end are where your expertise makes the difference.
Six modules. One practice task in each.
Work through them in order, or choose the question you want answered first. Use arrow keys to move between modules.
Module 1 of 6 · About 30 minutes
What is actually happening when AI writes a lesson plan?
A large language model has learned patterns from a very large amount of text. When you give it a prompt, it predicts a likely next word, then the next, until it has a response. It is very good at producing fluent, well-organised text. It does not look facts up in a textbook, and it does not know your class.
- Prediction, not lookup
- The model generates the most likely continuation of your prompt. That is why answers can sound certain even when they are wrong.
- Trained on the past
- Models learn from text collected up to a point in time. Recent syllabus changes, new textbooks or local exam formats may not be reflected.
- No memory of your classroom
- Unless you provide it, the model does not know your board, your learners, your timetable or what you taught yesterday.
- Different every time
- The same prompt can produce a different draft each time. Treat the output as one possible starting point, not the answer.
Ask an AI tool the same question three times, for example “Explain photosynthesis for Grade 6”. Compare the three drafts. What stays the same? What changes? Which would you actually use?
Module 2 of 6 · About 30 minutes
Why can a confident answer still be wrong?
Because the model is predicting text rather than checking facts, it can produce errors that look polished. Knowing the common failure patterns makes review much faster: you learn where to look first.
- Made-up detail
- Invented quotations, statistics, dates, page references or sources. Always check anything you would cite.
- Curriculum mismatch
- Content pitched at the wrong grade, using a different board’s terminology or skipping prerequisite knowledge.
- Calculation and reasoning slips
- Worked solutions with an arithmetic error in the middle, or answer keys that do not match the question.
- Bias and narrow examples
- Examples that only reflect one culture, gender or context, or language that is not inclusive of all your learners.
Generate a ten-question quiz on a topic you know well. Before using it, solve every question yourself and mark each one: correct, needs changing or remove. Note which failure pattern appeared most.
Module 3 of 6 · About 45 minutes
What does a good teaching prompt include?
The quality of a draft depends heavily on the context you give. A useful teaching prompt says who the learners are, what curriculum and topic you are teaching, what learners should be able to do by the end and what you want back.
- Learners and level
- Grade, age, prior knowledge and any support needs, without naming individual students.
- Curriculum and topic
- Board or framework, subject, chapter and the specific concept.
- A checkable objective
- What learners will be able to do, in words you could assess. “Explain why…” is better than “learn about…”.
- Format and constraints
- Lesson length, resources available, language level and the output you need: a lesson plan, three differentiated tasks, an exit ticket.
Rewrite a one-line prompt you have used before so that it includes all four kinds of context. Generate both versions and compare the drafts side by side.
Module 4 of 6 · About 45 minutes
How do I decide what to keep, change or remove?
Review is where teaching expertise matters most. A quick, consistent routine helps you catch errors and adapt the draft to your learners without starting from scratch.
- Accuracy
- Is every fact, definition, solution and answer key correct? Check against your textbook or a trusted source.
- Curriculum fit
- Does it match the objective, the grade and the terminology your board uses?
- Suitability
- Is the language, example and level right for these learners? Is anything inappropriate or exclusionary?
- Usefulness
- Does it actually help learners reach the objective in the time you have? What would you change after teaching it?
Take one AI-generated resource and annotate it with three colours: keep, change and remove. Write one sentence explaining your biggest change.
Module 5 of 6 · About 30 minutes
What should never go into an AI tool?
Teachers handle sensitive information every day. Using AI responsibly means sharing only what a task needs, following your school’s policy and being open with students and families about how AI supports your work.
- Share the minimum
- Describe needs in general terms, such as “three learners reading below grade level”, rather than entering names or personal details.
- Follow school policy
- Use tools your school has approved and follow its rules on data, accounts and consent.
- Keep assessment meaningful
- Design tasks that ask learners to explain their thinking, so assessment still reflects their understanding.
- Be open about AI use
- Be ready to explain to students and families how AI helps you prepare and that a teacher reviews everything.
Look at a prompt you might write about a class. Rewrite it so that no learner could be identified, while keeping the teaching need clear.
Module 6 of 6 · About 30 minutes
Which tasks are worth trying with AI first?
AI is most helpful on preparation tasks you do often, that take time and that you can review quickly. Decisions about individual learners, grades and relationships stay with you.
- Good first tasks
- Differentiated practice, question banks, alternative explanations, lesson outlines and exit tickets.
- Worth more care
- Assessment answer keys, feedback wording and anything that will be shared with families.
- Keep with the teacher
- Final grades, judgements about individual learners, safeguarding decisions and conversations that need empathy.
- Start small
- Choose one task for one class for one week. Review how it went before adding another.
List five preparation tasks from your week. For each, note how often you do it, how long it takes and how quickly you could check an AI draft. Pick one to try next week.
See what context changes.
Turn each kind of context on and off. Notice how the prompt becomes more specific and what a draft might get wrong without it.
Create a lesson on equivalent fractions for Grade 5 learners who can compare simple fractions but find equivalent fractions confusing.
What the draft may get wrong
- Curriculum and topic: Terminology and sequence may not match your board or textbook.
- A checkable objective: Without a clear objective, activities may be busy but unfocused.
- Format and constraints: You may get an essay when you needed a lesson outline you can teach from.
Practise the review habit.
This short draft sounds convincing. Read it as if you were about to teach it tomorrow.
Plants make their own food through a process called photosynthesis. They use sunlight, water and carbon dioxide from the air.
During photosynthesis, plants take in oxygen and give out carbon dioxide.
A 2021 study found that 82% of Grade 6 students misunderstand photosynthesis.
In the Calvin cycle, RuBisCO fixes CO₂ into 3-phosphoglycerate.
The food plants make is a sugar called glucose, which gives them energy to grow.
Can you find three problems?
Read the draft as if you were about to teach it. Look for errors of fact, invented detail and content that does not fit the learners.
- Accuracy
This is reversed. Plants take in carbon dioxide and release oxygen. A fluent sentence can still be wrong.
- Made-up detail
No source is given and the figure may be invented. Remove it, or find and cite a real study.
- Level and fit
Accurate, but far beyond a Grade 6 objective. It adds load without helping learners reach the goal.
The words you will hear.
Plain-language definitions to share with colleagues.
- Generative AI
- Software that creates new text, images or other content in response to a request.
- Large language model (LLM)
- A model trained on large amounts of text to predict likely words. It powers most AI writing tools.
- Prompt
- The instruction and context you give an AI tool.
- Context
- The information that makes a request specific: learners, curriculum, objective, format and constraints.
- Hallucination
- A confident but false or invented statement produced by an AI model.
- Grounding
- Asking AI to work from a source you provide, such as chapter notes, so you can check the draft against it.
- Bias
- A systematic pattern in the output, such as narrow examples or stereotypes, often inherited from training data.
- Human review
- The teacher’s check of accuracy, fit and suitability before anything reaches learners.
Take one lesson you teach next week.
Write a prompt for it with all four kinds of context. Generate a draft, then mark up what you would keep, change and remove. Teach the lesson and note one thing you would change next time. Keep this as the first piece of evidence if you plan to join the teacher certification.
Keep going.
AI-powered lesson planning
Turn one curriculum outcome into a connected lesson: objective, explanation, activity, practice and exit ticket.
Continue Essential for everyoneResponsible AI in the classroom
Practical habits for checking accuracy and bias, protecting student information and being open with students and families.
Continue ProgrammeTeacher AI certification
A structured programme that brings the Academy topics together and asks you to show thoughtful AI practice through a reviewed classroom portfolio.
ContinueA few things
you might be wondering.
Do I need any technical knowledge?
How long does it take?
Which AI tool should I use for the practice tasks?
Can I use this with my department?
Does this count towards the teacher certification?
Start with one lesson.
Build from there.
Bring the ideas into your next planning session, or talk to us about running AI foundations with your team.