How to ask an AI tutor better questions

A better question for an AI tutor tells it what you are trying to do, what you already know, and exactly where you got stuck, then asks for the kind of answer you need. Six habits do most of the work.

1. Give the reason you are asking.
2. State your level.
3. Point at the exact sentence, step or word that lost you.
4. Ask for an example, and sometimes a non-example.
5. Ask why, after you know what.
6. Ask what would change the answer.

It is what you would tell a human tutor in the first minute.

Why vague questions get vague answers

Anthropic's guide to prompting its Claude models says, "Think of Claude as a brilliant but new employee who lacks context on your norms and workflows." It suggests a test anyone can use. Show your prompt to a colleague with minimal context, and "if they'd be confused, Claude will be too." The same guide notes that giving the reason behind a request helps the model "deliver more targeted responses."

Google's guide for its Gemini models makes the same point with a router example. A general question about fixing a router gets a general answer. Adding one line of context, "The light on my Google Wifi router is yellow and blinking slowly," turns it into a question that can actually be answered.

Learning questions work the same way. "Explain photosynthesis" gets you the average answer to everyone who has ever asked it.

The six habits, with before and after

1. Give the reason

Before.

What is a p-value?

After.

I'm reading a news story that says a diet study was "statistically significant, p = 0.03." What is a p-value, and what does that number let me conclude about the diet, and what doesn't it?

The second version tells the tutor what you will do with the answer. That gives it a reason to skip the formula and focus on interpretation, which is what you need.

2. State your level

Before.

How does compound interest work?

After.

I'm comfortable with percentages but have never seen the compound interest formula. Explain how compound interest works using one example with real numbers over three years, before showing me any formula.

Level is often the most useful thing to add. Without it, the tutor has to guess, and an answer pitched at an imagined average reader is too slow for some people and too fast for others.

Toggle each detail and read the question as if you were the tutor. Keep the details that help explain where you need help.

3. Point at the exact thing

Before.

I don't understand this chapter on supply and demand.

After.

In the section on price ceilings, this sentence loses me. "A binding ceiling creates a shortage." Why would a maximum price cause a shortage instead of just making things cheaper?

A whole chapter is not a question. One sentence is. This is also where Learnfold's ask feature helps. You can select a passage in a lesson and ask "Why?", "How?", or for an example, and you do not have to retype or describe the sentence. Typing your own question works on the whole page and through "Where next?", so a question like the price ceiling one above can go there, quoting the sentence.

4. Ask for an example, and a non-example

Before.

What is a metaphor?

After.

Give me two examples of a metaphor and one sentence that looks like a metaphor but is actually a simile. Explain the difference in one line each.

Google's guide recommends always including examples in prompts, and using "specific and varied examples to help the model narrow its focus." For learning, the reverse is just as useful. Ask the tutor for examples, then ask for something that is almost an example but is not. The edge of a concept is where understanding gets tested.

5. Ask why

Once you know what something is, ask why it is true. Researchers call this elaborative interrogation, generating an explanation for why a stated fact is true. In John Dunlosky and colleagues' 2013 review of study techniques, it received a moderate utility rating, below practice testing and spaced practice but above rereading and highlighting. Dunlosky's summary for teachers frames it simply as encouraging students to ask "why?" while they study. The Learning Scientists suggest asking yourself "how these ideas work and why."

Before.

List the causes of World War I.

After.

Historians often list alliances as a cause of World War I. Why would alliances make a war more likely rather than less? Ask me to guess before you explain.

The Learning Scientists add a step people skip. After you come up with an explanation, "double-check your class materials" to make sure it is right. The same goes for the tutor's explanation.

6. Ask what would change the answer

Changing one condition and predicting the result separates understanding from memorizing.

Before.

Why does bread rise?

After.

Why does bread rise? And what would change if I used cold water, or added much more salt, or skipped kneading? Tell me which of those you're confident about and which depend on the recipe.

Asking what would change forces the explanation to show its moving parts. Asking which parts are certain invites the tutor to separate what it knows from what it is guessing.

Questions to keep handy

What would a common wrong answer to this look like, and why is it wrong?

Explain it again, but assume I already know X and skip that part.

Quiz me on this page with three questions. Don't give answers until I've tried.

What is the strongest objection to what you just said?

Which part of this explanation should I double-check, and where?

When this goes wrong

The answer is too long or too basic. Your level was missing or unclear. Say it, plus a length.

The tutor answered a different question. Quote the exact sentence or step you meant. In Learnfold, select it and ask from there, or type your question on the page with the sentence quoted. Some answers stay in the chat and some become pages of their own.

Your question had a wrong assumption in it. "Why is X true?" invites an explanation even if X is false, because the technique is built for facts that are already established. If you are not sure X is true, ask "Is it true that X?" first.

The answer sounds right and you cannot tell. That is the moment to check. Learnfold's own note applies to every AI tutor. AI makes mistakes, so treat it as a tutor, not a textbook, and check important facts against a reliable source. How to check an AI explanation for mistakes has a nine-step checklist.

You keep asking and never testing. Questions are half of studying. The other half is answering without help. How to study a textbook chapter with an AI tutor covers that side, and an AI tutor you can question is about doing both in one course.