Mastering AI Prompts: A Practical Guide for UK Secondary School Teachers

Large Language Models (LLMs) like ChatGPT, Claude, and Perplexity are rapidly becoming essential tools for educators looking to streamline tasks, create engaging resources, and support student learning. However, the adage "garbage in, garbage out" rings particularly true when working with AI. The quality of the output you receive depends heavily on the quality of your prompt.

Here are eight practical, classroom-ready tips for prompting LLMs effectively, which I have followed with much success in my teaching practice.

1. Be Clear and Specific in Your Instructions

Vague prompts lead to generic results. Instead, provide clear, detailed instructions which include the format, tone, style, or length you require.

Example:

• Weak prompt: "Create a revision guide on 'Macbeth'."

• Strong prompt: "Create a two-page revision guide on Macbeth for Year 11 GCSE students. Include bullet points summarising key themes (ambition, guilt), a brief analysis of Macbeth’s character, and three key quotes with explanations. The tone should be academic yet student-friendly."

Why this works: Specific instructions ensure the LLM delivers content that matches your needs without requiring excessive follow-up adjustments.

2. Provide Context and Background Information

AI performs better when it understands the purpose, audience, or intended use of the output.

Example: "Generate a 500-word introduction to the periodic table for Year 9 chemistry students. Focus on the historical development of the table, including contributions by Mendeleev, and explain how elements are organised by atomic number. Highlight how the periodic table is used to predict element properties and trends. This resource will be used as part of an introductory lesson on atomic structure."

Why this works: By specifying the audience (Year 9), the lesson stage (introduction), and the focus (historical context), the LLM can tailor its response more effectively.

3. Break Complex Requests into Smaller Steps

Instead of overwhelming the model with a multi-faceted prompt, break tasks into smaller, sequential prompts.

Example:

• Step 1: "Summarise the key causes of coastal erosion in 150 words for a Year 10 GCSE Geography student."

• Step 2: "Explain how human activity, such as coastal management strategies, can either accelerate or reduce erosion. Provide 3-4 bullet points with real-world examples."

• Step 3: "Now, write a paragraph that connects the impact of coastal erosion to challenges faced by coastal communities in the UK today."

Why this works: Breaking down a request into parts allows you to build on the AI’s previous responses and guide the output incrementally.

4. Be Specific About the Format

Clearly define the structure of the output you want. LLMs can produce text in various formats: tables, bullet points, essays, or even scripts for roleplays.

Example: "Create a bullet point list comparing the key events of World War I and World War II. Include the causes, significant battles, and outcomes for each."

Why this works: Specifying the desired format (bullet points) ensures the response is organised and easy to use in a classroom setting.

5. Set Constraints

Defining word count, tone, or topic boundaries helps refine the output.

Example: "Write a 250-word explanation of how to solve a quadratic equation using the quadratic formula. Focus on a step-by-step approach suitable for Year 10 GCSE Maths students. Include a worked example with clear explanations of each step, such as solving 2x² - 3x - 5 = 0."

Why this works: Constraints eliminate unnecessary tangents and ensure the response remains focused and concise.

6. Provide Examples

If you can, provide an example of what you want the AI to replicate. This could be a model answer, a resource style, or even just a starting point.

Example: "Here’s an example of how I want the resource to look:

• Key Topic: Present tense verbs

  • French: "Je mange une pomme."

  • Translation: "I am eating an apple."

  • Explanation: The verb "mange" is the first person singular form of "manger" (to eat) in the present tense. The subject pronoun "je" (I) indicates who is performing the action.

Now create three more entries for the topic of irregular verbs in French. Include the verb in its infinitive form, the conjugation for 'je' in the present tense, its English translation, and a sentence example."

7. Use Chain-of-Thought Reasoning

Encourage the AI to explain its thought process or break the task into logical steps.

Example: "Explain how to write a PEE (Point-Evidence-Explanation) paragraph for analysing religious concepts. Break this explanation into steps and provide an example using the concept of purgatory in Christianity."

Why this works: Asking for step-by-step reasoning makes the response clearer for students to follow.

8. Iterate and Refine

The first attempt isn’t always perfect. Don’t hesitate to tweak the prompt, ask follow-up questions, or request changes.

Example:

• First prompt: "Summarise Newton's three laws of motion for Year 9 students."

• Refinement: "That’s good, but could you simplify the language even more and focus on real-life examples like sports or car movement to make it relatable for students? Keep it under 200 words."

Why this works: Iterative prompting helps you refine the response until it meets your exact needs.

Final Thoughts

By following these best practices, we can make the most of large language models like ChatGPT. Whether you’re creating revision materials, drafting lesson content, or generating model answers, thoughtful, clear prompts will save time and result in high-quality outputs.