Prompt literacy is the new grammar: a practical ladder for ages 8 to 16

Asking a machine a good question turns out to be the same skill as thinking clearly. A rung-by-rung ladder for building prompt literacy from age 8 to 16, for parents and teachers who would rather teach the skill than ban the tool.

Grammar has always had a public-relations problem. Taught badly, it is a rulebook: underline the adverb, do not split the infinitive, lose a mark. Taught well, it is something else entirely - the load-bearing structure that lets a thought travel from one mind to another without collapsing on the way. Children who learn grammar in that second sense are not learning rules about language. They are learning to notice what they actually mean.

Prompt literacy is the same skill wearing new clothes. A prompt is a thought made explicit enough that a machine can act on it, and the machine is a brutally honest audience: it does exactly what you asked, which is frequently not what you meant. Every vague request answered with confident irrelevance is a small grammar lesson. The gap between "what I said" and "what I meant" has never been easier to see, or safer to practise closing.

That gap is worth teaching deliberately, because children will spend their working lives on one side of it or the other. Here is the ladder we use when thinking about ages 8 to 16 - not a curriculum, but a sequence of capabilities, each one resting on the last. Ages are rough. Children climb at their own pace, and plenty of adults are still on the bottom rungs.

Rung one, roughly ages 8 to 9: say what you actually want

The foundational skill is embarrassingly simple and genuinely hard: ask for the thing you want, one thing at a time.

At this age the work is noticing. A child asks "tell me about dogs" and gets an encyclopaedia entry when they wanted to know why their dog hides during thunderstorms. The teachable moment is not the answer; it is the mismatch. An adult sitting alongside can ask the golden question: "is that what you wanted?" And then: "what did you actually want to know?"

Practise these, and nothing more, at this stage:

• Asking one question rather than three tangled together.
• Saying who or what the question is really about ("my dog", not "dogs").
• Noticing out loud when the answer misses, and trying again with one change.

That last habit - change one thing, try again - is the seed of everything above it. It is also, not coincidentally, how scientists work.

Rung two, roughly ages 10 to 11: give the machine a "for"

The next capability is context. A question means different things depending on who is asking and why, and machines cannot infer the why unless told. So the child learns to attach a for to their requests: explain photosynthesis for someone who has never heard of it. Write instructions for my gran. Give me an example from football, because that is what I care about.

Two more habits belong on this rung:

• Constraints. "In three sentences." "Without giving me the answer." "Using words a Year 6 would know." Constraints teach children that they are in charge of the shape of the output, which quietly teaches them that the output is material to be shaped rather than an oracle to be believed.
• The second source. Anything that matters gets checked somewhere else - a book, a teacher, a parent, a different site. Not as a chore, but as a reflex, in the same way you look before crossing even a quiet road. At this age the check can be a game: the machine says X, the library book says Y, who is right and how do we find out?

Rung three, roughly ages 12 to 13: interrogate the answer

Somewhere around the start of secondary school, the relationship should change. The machine stops being a clever answerer and becomes a thing whose answers are produced - and anything produced can be produced badly.

The capabilities on this rung:

• Ask for the working. "How did you get that?" "What are you assuming?" A system's explanation of its own reasoning is not always faithful, and discovering that is itself a first-class lesson about these tools.
• Hunt confident nonsense. By this age a child should have personally caught the machine being wrong while sounding certain - ideally many times, ideally with delight. Teachers can engineer this: pick a topic the class knows deeply, generate an explanation, and set the class loose on it with red pens. Nothing inoculates against misplaced trust like the memory of catching the machine out yourself.
• Iterate on purpose. Not the toddler iteration of asking again louder, but deliberate revision: keep the request, change the audience; keep the audience, tighten the constraint; ask the machine to critique its own first attempt. Drafting, in other words. English teachers have taught this loop forever; the tool just makes each cycle cheap enough to do ten of them.

Rung four, roughly ages 14 to 16: stay the author

The top of the school-age ladder is not a prompting trick. It is a stance: the machine contributes, the human is the author, and the difference is non-negotiable.

Practically, this means learning to decompose. A GCSE-sized task - revise a topic, plan an essay, build a project - does not go into the machine whole. The student breaks it into parts, decides which parts are theirs alone (the argument, the judgement, the voice), and uses the tool on the rest: generating practice questions, stress-testing a plan, playing devil's advocate against a thesis, explaining a concept a third way when the first two did not land.

Students at this rung use the machine the way strong writers use a sharp editor: gratefully, frequently, and without ever letting the editor hold the pen. They can also do the one thing that proves authorship - defend the work out loud, unaided, when asked.

It is worth saying plainly: a student who reaches this rung has not learnt "prompting". They have learnt to make their thinking explicit, to specify audience and purpose, to revise deliberately, and to treat all claims as checkable. Those are the oldest goals in education. The tool is new; the literacy is not.

What to try this week

Pick the rung that matches your child or your class, and spend twenty minutes on it:

• Ages 8 to 9: side by side, let them ask an AI anything. After each answer, ask "is that what you wanted?" and let them try again with one change. Count the tries out loud; celebrate needing fewer.
• Ages 10 to 11: take one question and ask it three ways - plain, with a for, with a constraint. Compare the three answers together and talk about what changed and why.
• Ages 12 to 13: run a catch-it-out hunt. Choose something they know cold - their team, their game, a book they love - get the machine to write about it, and have them mark it like a teacher. Errors found are points scored.
• Ages 14 to 16: take one real piece of coursework preparation and split it on paper into "mine alone" and "tool can help". Then have them ask the machine to argue against their essay plan, and watch what that does to the plan.
• Teachers, any age: put a machine-written paragraph on the board without saying where it came from. Ask the class what is good, what is weak, and what they would check. You will not need to prepare a lesson on critical reading; the discussion is the lesson.