Teaching Prompt Library

Prompts for educators, plus the boundaries that matter when AI touches students' work.

The boundaries first

Never put a named student's work or personal information into a general AI tool. Anonymise it or use an institutionally approved tool. Student data is usually among the most protected categories there is.

Never let AI assign a grade. It can draft feedback and flag things to look at. The grade is a professional judgement with consequences for a young person, and it stays with you.

Never act on an AI-detection score alone. Detectors are unreliable in both directions and systematically flag non-native English writers. A score starts a conversation with a student; it never ends one.

Planning

Lesson plan from a specification

You are an experienced [SUBJECT] teacher working to [CURRICULUM/BOARD].

Topic: [TOPIC]
Year group: [YEAR], typical prior knowledge: [WHAT THEY KNOW]
Lesson length: [MINUTES]
Available: [BOARD / DEVICES / PRACTICAL EQUIPMENT]

Produce:
- One learning objective, phrased as what they will be able to do
- A five-minute hook that connects to something they already know
- The main sequence, timed
- One misconception likely to appear, and how to surface it
- An exit check that reveals whether they got it

Keep it practical. No "facilitate deeper engagement".

Differentiation

Here is my lesson core: [PASTE]

Produce three versions of the main task:
- Support: same objective, more scaffolding, smaller steps
- Core: as written
- Stretch: same content, requires them to justify or apply it

The objective must stay the same across all three. Do not water down
the concept for the support version — change the scaffolding, not the goal.

That constraint matters. Differentiation that lowers the objective is how gaps widen.

Misconception mapping

Topic: [TOPIC], year [YEAR].

What do students typically get wrong here?
For each misconception: what they believe, why it is intuitive,
what question would reveal it, and how to address it.

Feedback

Drafting feedback on anonymised work

Here is a student response to: "[THE QUESTION]"
Mark scheme: [PASTE]

[ANONYMISED RESPONSE]

Draft feedback:
- What they did well, specifically — not "good effort"
- The single most important thing to improve
- One concrete next step

Address the student directly. Under 80 words.
Do NOT suggest a grade or mark.

The whole-class pattern

Here are 25 anonymised responses to the same question: [PASTE]

What are the three most common errors?
For each: how many students, what the underlying misunderstanding is,
and what I should reteach.

Do not comment on individual students.

This is the highest-value use here: it turns a marking pile into a teaching decision.

Resources

Reading at the right level

Rewrite this passage for [YEAR GROUP] reading level.
Keep every technical term that is on the specification; explain each
on first use. Do not simplify the concept, only the language.

[PASTE]

Question generation

Generate 10 questions on [TOPIC] for [YEAR]:
- 3 recall
- 4 application
- 3 requiring justification or evaluation

For each, include the expected answer and one common wrong answer
with a note on what it reveals.
Match the style of [BOARD] questions.

Retrieval practice

Create a five-question retrieval starter mixing:
- 2 from last lesson
- 2 from last month
- 1 from last term

Topic sequence covered: [LIST]
Answers on a separate line so I can display the questions alone.

Admin

Parent communication

Draft a message to a parent about [SITUATION].
Tone: professional, warm, factual.
What I need them to know: [FACTS]
What I am asking: [IF ANYTHING]

No jargon. No blame. Under 120 words.
Do not include the student's name — I will add it.

Report comments

Here are my notes on a student: [ANONYMISED NOTES — effort, attainment,
specific examples]

Draft a report comment: what they can do, what to improve, one specific
next step. [WORD LIMIT].
Avoid generic praise. Every sentence should be something a parent could act on.

Always review these individually. A batch of AI-drafted reports that all sound the same is noticed immediately, by parents and by senior staff.

Teaching students to use AI

Worth doing deliberately rather than pretending it is not happening.

• Show them a hallucination. Ask for sources on something in your subject and verify them together. Nothing lands better.
• Compare AI output to a strong student answer. Ask what is missing. Usually: specificity, a position, evidence of thinking.
• Make the process visible. Ask for the prompts they used and what they changed. That is assessable and honest.
• Be explicit about what is permitted. Vague policies produce guessing, and the students who guess conservatively are disadvantaged.

Your boundaries

Approved tool at my institution
What I never put in
My stated policy for students
Who reviews AI-assisted reports