EazyTeach
AI in Education

Prompt Writing for Educators: A Practical Skill Worth an Hour

Teachers & Tutors7 min read

Most teachers who dismiss AI as producing generic rubbish are half right: it produced generic rubbish for them. The output of an AI assistant tracks the quality of the request closely, and the difference between a vague prompt and a well-built one is often the difference between unusable and excellent. The good news is that prompting is not a dark art. It is a small, learnable craft — closer to writing a clear brief for a new teaching assistant than to programming — and an hour of deliberate practice covers most of it.

The four elements of a working prompt

Role, context, constraints, and an example — most weak prompts are missing at least two of these.
  • Role: tell the AI what to be. 'You are an experienced Secondary 2 science teacher writing for students who struggle with reading' shifts vocabulary, tone, and depth all at once.
  • Context: the situation the output must fit — level, syllabus, what students already know, what happened last lesson. The AI knows nothing about your class unless you say it.
  • Constraints: format, length, difficulty, what to include and exclude. 'Five questions, no calculator required, answers as whole numbers, include a mark scheme' does more work than any adjective.
  • Example: paste one exemplar of what good looks like — a question in your house style, a feedback comment in your voice. A single example beats a paragraph of description almost every time.

Compare 'make a worksheet on fractions' with: 'You are a Primary 5 maths teacher. Write six word problems on adding unlike fractions, using Singaporean everyday contexts, difficulty rising from routine to challenging. Denominators no larger than 12. Include worked solutions. Here is an example in the style I want: ...' The second takes ninety seconds to write and saves twenty minutes of fixing.

Iterate — the first response is a draft

The most common usage mistake is treating the exchange as one round: prompt, disappointing response, give up. Experienced users treat the first output as an opening bid and refine in conversation. 'Make questions 4 and 5 harder.' 'Rewrite the explanation for a student who has not met algebra.' 'The tone is too formal — match the example I gave.' Each refinement takes seconds, and three rounds of it routinely turns a mediocre draft into a usable one. If a chat drifts badly off course, start fresh with a better prompt rather than wrestling a confused thread.

Common prompt mistakes

  • Vagueness: 'some questions on photosynthesis' delegates every decision — level, format, depth — to a system that knows nothing about your class. It will guess, and guess blandly.
  • Everything at once: one enormous prompt requesting a lesson plan, worksheet, quiz, and homework produces a shallow version of each. Sequence the requests; use the output of one as context for the next.
  • No verification step: a beautifully prompted response can still contain errors. Prompting well raises average quality; it does not remove the need to check facts and solve the questions yourself.
  • Accepting the first framing: if the output is organised wrongly for your purpose, say so — 'as a table', 'as a dialogue', 'as ten short items instead of one long passage'. Format is one sentence away.
  • Never saving what works: when a prompt produces exactly what you need, keep it. A personal file of five reliable prompts — worksheet generator, feedback drafter, explanation rewriter — quietly becomes one of your best teaching assets.

A one-hour practice plan

Take one real task you already need to do this week — a worksheet, a set of revision questions, a parent update. Write the lazy one-line prompt and look at the result. Then rebuild it with role, context, constraints, and an example, and compare. Spend the remaining time iterating: harder, simpler, different format, different voice. Most teachers find the skill clicks within a single session — and unlike most professional development, it starts repaying the hour immediately.

References & further reading
  1. White, J., et al. (2023). A prompt pattern catalog to enhance prompt engineering with ChatGPT. arXiv:2302.11382.
  2. Kasneci, E., et al. (2023). ChatGPT for good? On opportunities and challenges of large language models for education. Learning and Individual Differences.
  3. UNESCO — Guidance for generative AI in education and research (updated edition). UNESCO Publishing.