Everyday AISeptember 18, 2026

Turn one lesson into a 20-minute self-quiz with AI

A 20-minute study routine that turns one page of material into five self-test questions, grounded hints, and a final check without letting AI answer for you.

Beginner9 min read5 sources

You finish a page of a textbook and everything feels familiar. Then you close it and try to explain the idea in your own words — and discover that recognition is easier than recall.

Asking AI for another summary may make the material look clearer, but it does not tell you whether you can retrieve and use the idea yourself. A more useful pattern is to let AI ask while you answer.

With one page of study material and about 20 minutes, you can create a short self-test: write your own answer first, ask for one question at a time, request a hint only when needed, and check every correction against the original material.

This works for students and adults learning independently. Parents can also use the pattern with children, but AI should remain a support tool rather than a replacement teacher.

Start small and make the source clear

Choose one passage of roughly 300–700 words or one page of notes. That is a practical starting range, not a product limit. A full chapter makes questions drift and makes it harder to verify whether the assistant’s feedback is actually supported by the source.

Set aside about 20 minutes, keep a place to write your first answer, and use a normal ChatGPT conversation with Study mode if it is available to you. If you use another assistant, the fallback prompt below can still be useful, but behavior will vary across products.

Do not upload school records with names, health information, student contact details, unreleased exams, or internal documents. For minors, parents and teachers should consider age, school rules, and platform policies. UNESCO’s guidance on generative AI in education emphasizes human agency, data protection, and age-appropriate use.

First two minutes: answer before asking AI

Close the material and write three lines:

I think the main idea is:
Three terms I remember:
One thing I still cannot explain:

This is your baseline. Without it, you can tell only that the AI said a lot — not what changed in your own understanding.

Open the lesson again and mark one to three things you missed or remembered incorrectly. Keep the original answer unchanged so you can compare it with what you can explain later.

Give Study mode a specific job

OpenAI’s Help Center explains how to start Study mode in a normal ChatGPT conversation. The exact menu can change by app version, so use the current Help Center instructions if the option has moved.

Study mode is designed to ask questions, guide step by step, and check understanding instead of simply giving answers. OpenAI also notes that it can sometimes answer directly or make mistakes. Give it a short working agreement:

I want to test my understanding of the material below for 20 minutes.

GOAL:
[ONE THING I WANT TO BE ABLE TO EXPLAIN]

LEVEL:
[BEGINNER / SOME PRIOR KNOWLEDGE]

RULES:
1. Use only information inside <material>.
2. Ask one question at a time and wait for my answer.
3. Start with the main idea, then ask about detail,
   cause-and-effect, and one application.
4. Do not give the full answer before I try.
5. If I am wrong, give one short hint and let me try again.
6. Rate each answer: correct / partly correct / review needed.
7. Quote or point to the part of the material that supports feedback.
8. If the source is not enough, say:
   "The material is not enough to conclude this."
9. Stop after five questions and summarize:
   understood / needs review / one idea I should explain again.

<material>
[PASTE THE STUDY PASSAGE]
</material>

What a useful exchange looks like

The following material is synthetic and created only for illustration:

Water in rivers, lakes and oceans absorbs heat and evaporates.
Water vapor rises, cools and condenses into small droplets that
form clouds. When droplets become large enough, they fall as rain.
Some water enters the ground and some flows back into rivers,
lakes and oceans, where the cycle continues.

A useful exchange might look like this:

AI: Close the material and explain the water cycle in three sentences.

Learner: Water from the ocean becomes clouds. Clouds make rain.
The water goes back to the ocean.

AI: Partly correct. You have the main path. The material also names
what causes evaporation and what causes condensation. Add those two
causes in your own words.

This dialogue is an editorial example, not a recorded output from a specific model run. The important behavior is that the assistant does not replace the learner’s answer. It points to two missing links and asks the learner to supply them.

A bad version would immediately expand into extra terminology and facts that are not in the supplied passage. Those facts might be correct elsewhere, but they would weaken this particular test. The goal is to find out what you understand from the material you are actually studying.

Last five minutes: verify instead of chasing a score

After question five, do not ask, “What score did I get?” Make a simple table:

Question	Answered immediately?	Needed a hint?	Feedback grounded in source?
1	Yes/No	Yes/No	Yes/No
2	Yes/No	Yes/No	Yes/No
3	Yes/No	Yes/No	Yes/No
4	Yes/No	Yes/No	Yes/No
5	Yes/No	Yes/No	Yes/No

Choose one question that needed a hint. Close both the chat and the source, then answer it again. If you still need to look at the answer, mark it “review needed” rather than treating the session as complete.

A useful session leaves you with your baseline, one-question-at-a-time practice, feedback tied to the source, and at least one idea you can now explain without AI. There is no universal percentage that proves you have learned a lesson; course and teacher requirements still matter.

When AI breaks the self-test

It gives the answer immediately. Say: “Do not give the full answer. Ask one question or give a hint of no more than 15 words.” If it keeps answering for you, start a new chat.

The questions are too easy. Ask for a cause-and-effect question or an application scenario, while still limiting the assistant to the source.

It adds outside knowledge. Ask: “Quote the sentence in that supports that feedback. If it is not there, withdraw the claim.” If no support exists, remove that question from your study result.

It misreads an image or PDF. Paste the relevant text directly and identify the page or section. OpenAI’s Study mode guidance also recommends clearer images or pasted text when content is missed.

You cannot find Study mode. Check that you are signed in and using a normal chat, then consult the current Help Center instructions. If it is still unavailable, use the prompt above in a standard conversation or write five questions yourself from headings and key ideas.

You become dependent on hints. In the next session, allow at most two hints. AI is helpful when it exposes a gap; it is counterproductive when it fills every gap with a ready-made answer.

This workflow has not been run here as a controlled test on one fixed account, app version, and model. Try it first with old material whose correct answers you or a teacher already know, and check whether the assistant waits, grounds feedback in the source, and avoids answering for you.

When not to use this method

Do not use AI on graded work when your teacher or institution does not allow it. Do not upload unreleased exam questions or restricted material to an external service. For laboratory work, pronunciation, performance skills, or other tasks that need physical practice or expert observation, text chat is not a substitute for the real activity.

OpenAI itself says Study mode does not replace teachers, course materials, or academic requirements. Its student guidance also encourages transparency about AI use and checking information. These are still vendor-provided materials; your institution’s rules and your teacher’s judgment take priority.

A good self-quiz does not end with a polished summary. It ends with something you could not explain at the start but can now explain in your own words.