A Socratic AI tutor prompt works when it slows the model down. The AI should not dump a full answer immediately. It should ask a question, wait for your attempt, give a hint, and only explain after you commit.
Use this when
- You keep rereading notes but cannot explain the concept out loud.
- You want the AI to tutor you without doing the work for you.
- You need hints and follow-up questions before final explanations.
- You want to identify the exact step where your reasoning breaks.
What the prompt should produce
- A one-question-at-a-time tutoring loop.
- Hints before answers and corrections after your attempt.
- A short weak-spot summary after each round.
- Follow-up questions that connect definitions to examples.