Prompt Engineering for Teachers: Beyond 'Write Me a Lesson Plan'

The most common AI prompt I see teachers write is some version of "Write me a lesson plan on [topic] for [year group]." It works in the sense that the AI produces something. But what it produces is generic, surface-level, and almost always needs significant reworking before it is usable. Teachers try it, find the output disappointing, and conclude that AI is overhyped.

The problem is not the AI. The problem is the prompt. And the solution is not a list of better prompts to copy; it is understanding the principles that make any prompt effective. Prompt engineering is not a technical skill. It is a communication skill. And teachers, who spend their professional lives crafting explanations, questions, and instructions for diverse audiences, are actually better positioned to master it than most people.

Why Most Teacher Prompts Fail

When a teacher writes "Write me a lesson plan on photosynthesis for Year 9," they are giving the AI almost no useful information. The AI does not know the ability range of the class. It does not know the prior knowledge students have. It does not know the assessment objectives being targeted. It does not know the school's preferred lesson structure. It does not know whether the teacher wants a practical lesson, a discussion-based lesson, or a direct instruction lesson. It does not know the time allocation.

So the AI makes assumptions about all of these things, and the result is a lesson plan that fits no specific context. This is not AI failure; it is prompt failure. The AI is doing exactly what it was asked to do: write a generic lesson plan with no context.

The principle is simple: the more specific and contextual your prompt, the more useful the output. But "be more specific" is not particularly helpful advice on its own. Let me break it down into practical strategies.

Strategy 1: Define the Role

Tell the AI who it is. This seems trivial, but it fundamentally shapes the output. "You are an experienced secondary science teacher at a British curriculum international school" produces very different output from the default. The AI draws on different patterns, uses different vocabulary, and makes different assumptions about context.

I recommend starting every educational prompt with a role definition that includes the subject, the phase (primary, secondary, sixth form), the curriculum context, and any relevant specialisms. This is your baseline context; set it once and it informs everything that follows.

Strategy 2: Provide the Context the AI Cannot Guess

The AI does not know your class. You do. Bridge that gap explicitly.

Instead of: "Write a lesson plan on photosynthesis for Year 9."

Try: "I am planning a 60-minute lesson on photosynthesis for a mixed-ability Year 9 class of 28 students. They have already covered cell structure and basic plant biology. The class includes 4 students with EAL needs and 2 students who are working above age-related expectations. The lesson needs to cover the light-dependent reactions and should include a practical component. We are working towards IGCSE Biology learning objective 2.18."

The difference in output quality is dramatic. You have given the AI enough context to produce something genuinely tailored, rather than forcing it to guess.

Strategy 3: Specify the Output Format

Do not let the AI decide how to structure its response. Tell it.

"Present this as follows: learning objectives, starter activity (5 minutes), main activity (30 minutes), practical component (15 minutes), plenary (10 minutes). For each section, include the activity description, key questions to ask, and differentiation notes for lower and higher attainers."

This level of structure gives you output that maps directly onto your school's lesson planning template, saving you the time of reformatting generic content.

Strategy 4: Use the Chain-of-Thought Approach

For complex tasks, do not ask the AI to produce the final output in one go. Break the task into steps.

Step 1: "Based on IGCSE Biology syllabus point 2.18, identify the three most important concepts students need to understand about light-dependent reactions."

Step 2: "For each of these three concepts, suggest a common misconception that Year 9 students typically hold."

Step 3: "Design a starter activity that surfaces these misconceptions through a multiple-choice diagnostic quiz."

Step 4: "Now design the main teaching sequence that addresses each misconception in turn."

Each step builds on the last, and you can course-correct at each stage. If the AI identifies the wrong misconceptions in Step 2, you fix it before it builds a lesson around them. This is far more effective than hoping a single prompt produces a perfect lesson from scratch.

Strategy 5: Ask for Alternatives, Not Answers

One of the most powerful but underused prompting strategies is asking the AI to generate options rather than a single output.

"Give me three different approaches to teaching the light-dependent reactions to a mixed-ability Year 9 class. For each approach, explain the pedagogical rationale; the main activity; and the potential challenges."

This shifts the AI from a content generator to a thinking partner. You are not asking it to do your job; you are asking it to expand your options so you can make a better-informed decision. This is the kind of AI use that genuinely enhances professional practice rather than replacing it.

Strategy 6: Build Iterative Refinement into Your Workflow

Your first prompt will almost never produce perfect output. This is normal and expected. The skill is not writing perfect prompts; it is refining output through conversation.

After the AI generates a lesson plan, respond with: "The starter activity is too complex for the lower-ability students in this class. Simplify it and add a scaffolded version with sentence starters." Or: "The practical component needs a risk assessment summary. Add one." Or: "The plenary assumes students can articulate their understanding verbally. Add an alternative for students who express themselves better in writing."

Each refinement makes the output more tailored to your specific context. The AI learns from each instruction within the conversation; by the third or fourth refinement you typically have something genuinely useful; something that would have taken significantly longer to create from scratch.

Beyond Lesson Planning

Lesson planning is the most common use case, but the same principles apply across professional practice.

Assessment design: "Design a 30-mark assessment on photosynthesis for Year 9 IGCSE Biology. Include questions at AO1 (recall), AO2 (application), and AO3 (analysis) levels, with the marks weighted 40/40/20. Include a mark scheme with acceptable responses and common errors."

Feedback writing: "Here is a student's response to a 6-mark question on photosynthesis: [paste student work]. Write feedback that identifies two specific strengths, one area for improvement, and a follow-up question that would deepen the student's understanding. Use language appropriate for a Year 9 student."

Differentiation: "I need to adapt this reading passage on photosynthesis for three ability levels: below expected (reading age 9-10), at expected (reading age 12-13), and above expected (reading age 15+). Keep the scientific accuracy consistent across all three versions but adjust the vocabulary and sentence complexity."

Parent communication: "Draft an email to parents explaining how our Year 9 science class will be using AI tools this term. The tone should be reassuring and transparent. Address likely concerns about assessment integrity and data protection. Keep it under 300 words."

In each case, the same principles apply: define the role, provide the context, specify the format, and refine iteratively.

The Bigger Point

Prompt engineering is not about memorising templates. It is about developing the habit of communicating precisely with AI systems: understanding what information they need, what assumptions they make when information is missing, and how to iteratively refine output to meet your specific needs.

Teachers who develop this skill do not just become better at using AI. They become better at articulating their pedagogical thinking; crafting a good prompt requires you to be explicit about your objectives, your students' needs, your differentiation strategies, and your assessment criteria in a way that unstructured planning does not.

That, ultimately, is the promise of AI in education: not that it does the thinking for you, but that it requires you to think more clearly about what you want. And for teachers, thinking clearly about what you want for your students is the whole game.