When you first give students access to an AI tool, half of them ask: "What do I even say to it?" They know the tool exists, they know it can answer questions, but they freeze. The prompt feels too open. They don't know the rules, so they either ask nothing at all or ask something so vague the AI gives back vague nonsense.
This is actually a thinking problem, not a tool problem. And you can teach it.
What's hard about prompting
Students who learn to write essays, answers to prompts, and emails all know the basic rule: *know your audience*. Prompts to AI should work the same way. But AI feels like a black box. They don't have intuition for what it needs to produce something useful.
So they:
- Ask it to do everything at once ("help me with my project")
- Ask it to do something they haven't defined yet ("write a story")
- Ask vague questions and then blame the tool for vague answers ("Is the Civil War important?")
This is the same mistake that happens with human writing: they haven't thought through what they want, so the output is muddy.
The teaching move: one question at a time
The skill to teach is breaking a big, fuzzy request into specific, answerable questions.
Start with something they *already do* in writing class:
"When you get an essay prompt, you don't just start writing. You break it down. What's it asking? Who's my audience? What does success look like? Same thing with AI. If you can't ask a human the question clearly, the AI won't understand it either."
Then give them a worked example:
Fuzzy: "Help me understand the French Revolution"
Specific: "Explain in 3-4 sentences why the French Revolution started. What were the main causes? I'm trying to understand the difference between grievances the people had and the actual trigger for violence."
Walk through that with them. "The first one is too big. Where do I even start? The second one narrows it down: I'm asking for a specific length, I'm asking for a specific frame (causes, then trigger), and I'm saying why (so the AI knows I need clarity, not just facts)."
The practice structure
Give students a scenario and have them turn a fuzzy request into a specific prompt. Don't ask them to use AI yet. Just practice the prompting.
Scenario: "You want to use AI to help you plan a lesson you're teaching someone else."
Fuzzy: "How do I plan a lesson?"
Specific: "I need to teach [topic] to [audience] in [time]. I want students to learn [what]. What should I include in the lesson plan? What's essential and what can wait?"
This is a real skill. Their prompts get better as they do it more.
The conversation after they get an answer
Here's where you catch the thinking:
"You asked the AI for X and got Y. Is Y actually what you needed? If not, what's missing? What would you ask differently next time?"
This teaches them that a bad answer isn't the AI's fault; it's usually because the question wasn't clear enough.
Sometimes they'll say: "I asked for a summary and it gave me the whole thing." That's a chance to coach: "What if you asked it to summarize in 2 sentences? Or to summarize just the most important causes?"
It's the same metacognitive move they learn in reading: *what did the text actually say vs. what did you expect?* Here it's: *what did the AI actually give you vs. what did you ask for?*
Where this gets easier
Our [learning objectives tool](/learning-objectives) helps you define exactly what students should be able to do with a topic, and our [student onboarding guide](/student-onboarding) walks students through setting up their own learning process. Both are natural pairs with prompt-writing practice, because clear objectives make for clear prompts.
The test of a student who's learned to prompt well
They get an answer from an AI and can tell you:
- Whether it actually answers the question they asked
- What they'd ask differently if the answer missed the mark
- What the AI seemed to understand and what it misunderstood
They're not blaming the tool. They're diagnosing their own asking.
That's the move. That's the skill worth teaching.
