Generic prompts ("help me plan a maths lesson") produce generic results. The more specific your prompt, the more useful the output — here are tested starting points by subject for AI in primary schools.
Literacy
"Generate five discussion questions about [book/chapter] suitable for Year 4, ranging from literal to inferential." Add "include one question specifically about the author's word choice" for a stronger stretch question.
Maths
"Create 8 fluency questions and 2 reasoning questions on [topic] for Year 3, with an answer key." Specify the exact strategy pupils have been taught (e.g. "using the column method") to avoid mismatched approaches.
Science
"Suggest three simple, safe classroom demonstrations for teaching [topic] to Year 2, using only common classroom materials." Always check the safety of any suggested demonstration yourself before running it.
Foundation subjects
"Draft a knowledge organiser structure for a Year 5 unit on [topic], covering key vocabulary, dates/facts, and a summary." Ask for a second version pitched at a lower reading age if you need a differentiated copy too.
Why specificity matters so much
The difference between a mediocre and an excellent AI output is almost always specificity in the prompt — year group, time constraints, exact method taught, and the format you want the answer in. Spend the extra fifteen seconds adding this detail and the output usually needs far less editing afterwards.
See our full Prompt Library for a larger set organised by role.
Key takeaways
- Specific prompts (year group, method, format) beat generic ones every time.
- Always check safety and accuracy of AI-suggested demonstrations or maths methods.
- Build subject-specific prompts into your personal library as you find ones that work.