Classroom practice
AI for Educators
Where AI genuinely helps in teaching: planning, scaffolding, assessment preparation, AI literacy and workload. Practical starting points for primary, secondary, FE, adult learning and support staff — with every output checked by you.
AI drafts; you decide. Every example on this page keeps the professional judgement — and the final decision about a learner — with the human being.
Planning
AI is at its best as a planning assistant: it can produce a first-draft lesson structure, five alternative ways to open a topic, or a cover lesson for a non-specialist in seconds. It has never met your learners, so treat every plan as a draft for you to adapt — timing, pitch and context are yours to fix.
- Paste a learning outcome and level into an approved tool and ask for a draft plan.
- Check the subject content against a source you trust.
- Adjust timings and pitch for your actual group.
- Ask the tool to critique its own plan — what’s unrealistic, what assumes prior knowledge?
Example: Level 1 vocational
A vocational lecturer asks for a 60-minute lesson on health and safety in a workshop. The draft arrives with a starter, input, practical and plenary — and a step that assumes learners can read a risk assessment unaided. The lecturer knows half the group can’t yet, and adds a paired reading step before using it.
Ready-made planning prompts: prompts 1–6 in the Toolkit, or the ten fuller lesson-planning templates.
Differentiation and scaffolding
AI can rewrite a text at a different reading level, build a support/core/stretch version of one task, produce sentence starters and writing frames, or fade a worked example into independent practice. The key discipline: scaffolding changes the route, not the destination — and the original full-level material stays available to every learner.
Example: Entry Level English
A support assistant pastes an exam-style instruction (with identifying details removed) and asks for numbered steps, one action per step, maximum twelve words each, ending with what a finished task looks like. She checks each step against the original before the lesson — one step had quietly dropped a requirement, and she put it back.
See SEND, Inclusion & Accessibility for barrier-led guidance, and prompts 7–12 in the Toolkit.
Assessment preparation (not assessment decisions)
AI can draft multiple-choice questions with plausible distractors, build a plain-English marking checklist, or suggest low-stakes checks that need no marking. What it must never do is become the sole or primary means of marking, or make final decisions about a learner’s progress. Awarding-body rules on AI use in assessments apply — see the Research & Guidance page for current JCQ guidance.
For any assessed work, check your awarding body’s current AI rules before using AI anywhere in the process — including feedback drafting. Where AI use is permitted, learners may need to acknowledge it. AI misuse in assessments is treated as malpractice.
Feedback drafting that you edit: prompt 15. Quiz and checklist prompts: 13–18.
AI literacy for learners
Learners need to understand AI as much as staff do: when it helps, when it’s wrong, and what the rules are for their courses. The Toolkit’s AI-literacy prompts (27–33) build short activities — comparing an AI answer with a textbook, fact-checking deliberately wrong output, discussing when AI use is acceptable in coursework.
Example: Level 2 study skills
A tutor gives groups an AI-written paragraph about local employment statistics containing three deliberate errors. The task: find them, using official sources. The room learns more about checking sources in twenty minutes than in a term of warnings.
Workload: the honest position
AI can genuinely take the sting out of repetitive drafting — reminders, summaries of your own notes, template emails, reformatting. It cannot mark for you, know your learners, or take accountability for decisions. Start with the tasks you would happily hand to a competent stranger with a good dictionary, and keep everything that needs your judgement.
Workload prompts: 39–44 in the Toolkit.
Common mistakes
- Pasting learner names or work into a general tool — never enter identifiable learner data.
- Using output unchecked because it looked right once.
- Assuming one tool suits every learner — offer options.
- Letting AI write feedback you wouldn’t say yourself.
- Skipping the awarding-body rules on assessed work.
Safeguarding and ethics
Use organisation-approved tools only; know your organisation’s rules and who handles data protection; know what to do if a learner meets upsetting content through AI. The full Safe-Use Checklist covers this in seven sections.
Downloads for this page
Page last reviewed: 7 November 2026 · Next review: by May 2027 · See Research, Guidance & Evidence for dated official sources.