ChatGPT Learning Mode System Prompt: A Reusable Tutor Template

A ChatGPT learning mode system prompt turns a general assistant into a structured tutor: one that asks before it explains, checks before it moves on, and slows down when you are stuck. The default ChatGPT behavior is to answer; a tutoring prompt shifts it toward questioning, which is how people actually learn something rather than just read it.

You can set this up once, save it, and reuse it across subjects. Tools like fenbs store the approved prompt alongside the lessons from testing it, so you pick up the working version rather than rewriting it each semester.

What a tutoring system prompt does differently

Without a tutoring prompt, ChatGPT tends to answer the surface question completely and move on. That is useful for looking things up; it is not useful for building understanding. A tutoring prompt changes three things:

• It asks the student what they already know before explaining anything.
• It uses questions and examples rather than definitions as the main explanation method.
• It checks understanding before introducing the next concept.

These changes shift the session from passive reading to active recall, which research in learning science consistently identifies as more effective for retention. The prompt does not make ChatGPT smarter; it makes the interaction structure better.

The reusable tutoring prompt

Copy and adapt the following. Replace the bracketed variables with the subject, level, and session goal before starting.

You are a patient and precise tutor helping a student learn [SUBJECT] at [LEVEL: beginner / intermediate / advanced].

Rules:
1. Before explaining any concept, ask the student what they already know or believe about it.
2. Use questions, examples, and analogies as your main tools. Give definitions only after the student has tried to describe the concept in their own words.
3. After explaining any concept, ask one check question before moving on. Wait for the answer.
4. If the student answers incorrectly, do not say "wrong." Instead, ask a narrower question that leads them toward the right answer.
5. If the student says they are stuck or confused, break the concept into a smaller piece and start there.
6. Limit each response to three paragraphs or fewer. Do not cover more than one concept at a time.
7. Keep track of which concepts have been covered and confirmed. Before the session ends, summarize what was learned and what still needs review.

Today's goal: [STATE THE SESSION GOAL — e.g., "understand what a derivative represents geometrically"].

How to adapt the template for your subject

Level matters more than subject. A beginner studying contract law needs different pacing than a third-year law student reviewing the same concepts. Be specific about the level and the background you are bringing, not just the topic.

The session goal keeps the tutor focused. Without it, a chatbot tends to wander across a topic. A goal like "understand how a capacitor stores charge, not how to calculate it" is precise enough to keep a 30-minute session on track.

You can also add domain-specific rules. For a coding subject: "When I write code, point out one problem at a time and ask me to fix it before showing me the answer." For a language subject: "Correct grammar errors by asking me to try again rather than correcting them directly." These additions make the behavior consistent across sessions.

What the prompt does not handle

This prompt makes the session structure better; it does not guarantee factual accuracy. ChatGPT can give plausible but incorrect explanations, especially in technical or specialized subjects. Check important claims against a textbook or primary source, particularly in law, medicine, or engineering.

The prompt also does not track progress across separate conversations. Each new chat starts fresh. If you want to continue from where a previous session ended, you need to paste a summary of what was covered, or use a tool that saves session notes for your next AI conversation.

For subjects that require real-time feedback, such as language pronunciation or physical technique, a text-based tutor cannot replace a human teacher or a purpose-built app. The tutoring prompt handles understanding and recall well; it handles anything that requires listening, correction of physical form, or demonstration poorly. Know which parts of your subject fall outside its scope before relying on the template for exam preparation.

Saving and improving the prompt over time

The first session will reveal gaps. You may find the check questions come too frequently, or that the three-paragraph limit cuts off a necessary explanation. Note what did not work after each session and update the template. After three or four sessions, you will have a version tuned to how you actually learn.

Save the lessons from testing alongside the prompt. A note that says "adding 'give one concrete example before asking a check question' significantly improved sessions on abstract concepts" is more useful than the memory that something was changed. The guide on system prompt examples shows how to structure similar entries.

Pull the subject-specific rules out of one session's template once you find they apply broadly. An instruction like "when I answer incorrectly, show me the error in my reasoning before giving the correct answer" applies to any subject where understanding, not just recall, is the goal. A standing set of these rules, separate from the session-goal variable, means each new session starts with the accumulated improvements from earlier ones.

Consider sharing the template with a study partner working through the same material. A second tester surfaces gaps that only appear with a different learning style or pace. Two rounds of real testing catch most common problems before you rely on the template for exam preparation or important work.

How to do this with fenbs

1. Create a free account and open a project for the subject you are studying.
2. Add the tutoring prompt to Skills and prompts. Include the subject, level, session goal format, and any subject-specific rules you added.
3. After each session, record what worked and what did not in Lessons learned, linked to the prompt.
4. Update the prompt based on each lesson. Note the change and the date so you can track which version produced which improvements.
5. Connect your AI assistant over MCP from Integrations so it reads the current prompt and lessons at the start of each session.