Tell ChatGPT what you want to understand, what you already know, and where you get stuck. Ask for a plain explanation, one worked example, and the conditions that matter. Then explain the idea back in your own words and try a new example to find what is still unclear.
What you need before you start
A topic or passage is hard to understand and you want a clear explanation you can use.
Bring: The question, what you already understand, the part that confuses you, and what you want to do next.
Start with these steps
- Name the exact question and what you already know.
- Ask for easy words, a clear explanation, and one worked example.
- Ask what conditions and limits matter.
- Explain the idea back without reading.
- Try a changed example and ask for a hint where you get stuck.
A prompt you can adapt
Explain [topic] for a learner who knows [background] and needs it for [purpose]. Start with a map, then give the mechanism, one worked example and one boundary case. Separate established facts, analogies and uncertainty. Ask me one check question at the end.
Replace every bracket with real information. Remove fields that do not affect your task.
What this looks like in practice
Teaching example: to understand a weighted average, start with two test scores that count for different amounts of a final grade. Calculate the result, then change the weights to see why an ordinary average gives the wrong answer.
Name the thing you want to understand and why
A broad request such as “explain economics” gives the model too many possible starting points. Choose the question that is stopping you. Perhaps you want to know why a loan payment changes, how a weighted grade works, or what a computer error means. Tell ChatGPT what you need to do after reading. The explanation can then focus on the part that helps you act.
Say what you already understand in ordinary words. You do not need to name a school level. “I can work out a normal average, but I do not understand why some scores count more” is useful context. If you are starting from nothing, say that. If you have a passage that confused you, paste it and identify the sentence where you lost the thread.
Ask for easy language without asking it to remove the information. A good request is: “Use familiar words. Define any necessary technical term the first time you use it. Keep the conditions and show one example.” You can understand a difficult subject through simple language. The model should not replace the subject with a cheerful but empty summary.
Ask for layers that build on each other
Start with a short answer to the main question. Follow it with the mechanism: what changes, why it changes, and what causes the result. Then ask for an example with actual steps. This order lets you see the point before meeting the details. A definition alone may be accurate while still leaving you unable to use the idea.
For a term you do not know, ask for its meaning in the current sentence. A general dictionary-style definition can add more unfamiliar words. If the topic is a computer program, the relevant meaning of a word may differ from its everyday meaning. Ask what the term does in the example you are discussing.
You can control the next layer with a follow-up. “Explain the second step more slowly” is better than requesting the entire explanation again. If you understand the basic example, ask what changes in a harder case. This keeps progress connected to your actual gap instead of giving you repeated introductions.
Work through a weighted average with named quantities
Here is a teaching example with invented grades. A course uses a quiz worth twenty-five percent of the final grade and an exam worth seventy-five percent. You score eighty on the quiz and sixty on the exam. A normal average gives seventy, but it treats both scores as equally important. The course does not do that: the exam counts three times as much.
Multiply the quiz score by its share: eighty times 0.25 is twenty. Multiply the exam score by its share: sixty times 0.75 is forty-five. Add the two contributions to get sixty-five. The weights add to one, so the result is already on the same scale as the original grades. Each multiplication answers how much that score contributes to the final grade.
Ask ChatGPT to show the names beside the numbers. “80 × 0.25” is easier to understand when it says quiz score times quiz share. If the result comes out higher than eighty or lower than sixty with positive weights totaling one, check the calculation. A weighted average of these two scores should stay between them.
| Assessment | Score | Share of final grade | Contribution |
|---|---|---|---|
| Quiz | 80 | 25% | 20 |
| Exam | 60 | 75% | 45 |
| Final grade | 100% | 65 |
Final grade = (80 × 0.25) + (60 × 0.75) = 65
Ordinary average = (80 + 60) / 2 = 70
The ordinary average uses equal weights, which this course does not.Use an analogy, then return to the real thing
An analogy compares an unfamiliar idea with something familiar. It can give you a useful first picture. For a weighted average, you might imagine one vote counting once and another vote counting three times. That shows unequal influence. It does not by itself explain every calculation, so ask where the comparison stops being helpful.
Every analogy leaves something out. A story about water flowing through pipes may help explain a relationship, but it should not become a substitute for the actual quantities and conditions. Ask ChatGPT to name the parts that match and the parts that do not. Then return to a real example from the topic.
If the analogy makes you more confused, ask for a different form. A small table, a labelled drawing, or a step-by-step calculation may work better. You do not need to keep an analogy simply because it sounds clever. The useful test is whether you can explain the actual idea more clearly after using it.
Keep the exceptions and limits in plain sight
An easy explanation can become wrong if it drops a condition. Ask “When does this rule apply, and when would it fail?” For the grade example, the weights need to describe the whole grade if you want a final course result. If there is another assignment not included, sixty-five is only the result for the pieces you supplied.
If weights are written as one and three rather than 0.25 and 0.75, multiply the scores by one and three, add them, and divide by four. The division matters because the weights do not add to one. Ask the model to show why those two ways describe the same proportions. You should not copy a formula without knowing what its parts mean.
For real-world questions, also ask what information is missing. An explanation may describe a general mechanism without predicting a particular outcome. Knowing how interest works does not establish the cost of a specific loan without its terms. Keep the general idea separate from the facts needed for your own case.
Explain the idea back without reading
Close or cover the answer and explain the idea in your own words. You might say, “A weighted average gives some values more influence. I multiply each value by its share, then add the contributions.” Ask ChatGPT to identify a missing condition or a mistaken link. Tell it not to praise the answer before checking it.
If your explanation uses a term you cannot define, pause there. Ask for one small example of that term. You may recognize a good explanation without being able to produce one. Explaining back exposes that gap kindly and quickly. It does not need to sound polished.
Ask a new question with different numbers or a different setting. For the grade example, try a score of fifty on a component worth forty percent and a score of ninety on one worth sixty percent. The result is seventy-four. Work it out yourself before seeing the answer. A new example tests understanding better than repeating the calculation you just read.
Repair an explanation that is too hard or too simple
If the answer contains too many unfamiliar words, name the first one and ask for the passage to be rewritten around it. “Explain denominator in this example” is a focused request. “Use shorter words everywhere” may replace necessary terms while leaving the difficult relationship unexplained. The relationship is the part you need to understand.
If the answer is too simple, ask for the missing mechanism, condition, or consequence. For example, ask why doubling all weights leaves a normalized weighted average unchanged. The sum of the weighted values doubles, and the total weight doubles, so the ratio stays the same. That adds depth without needing fancy language.
If the model contradicts itself, quote the two statements and ask it to resolve the difference using a worked example. Check important facts at a suitable source. ChatGPT can explain clearly and still make an error. For a consequential decision, the explanation should help you ask better questions and check the facts, rather than act as the only authority.
Build a small note you can use later
End with a note in your own words: the main idea, the example, the condition that matters, and one question you still have. Keep it short enough to revisit. A copied page may feel complete, but a small note you understand is easier to use when you meet the topic again.
Return later and try a different example before rereading. If you get stuck, use the note to identify the missing step. You can ask ChatGPT for a hint rather than the full answer. The aim is to become less dependent on the explanation as you learn.
You have made useful progress when you can explain why the method works, apply it to a changed case, and recognize when more information is needed. You do not need to know every detail of a broad subject in one session. Keep building from the specific question you can now answer.
Check the result before you use it
- Can you explain the main idea in your own words?
- Can you apply it to a different example?
- Do you know when the rule needs more information?
Watch for: Understanding the example you just read is different from solving a new one yourself.