Verified against ChatGPT · 2026-08-12
Run a strict Socratic tutoring session where the AI is only allowed to ask questions, never state facts
Sets up a tutoring conversation with a hard rule that the AI can only respond in questions, forcing you to construct the reasoning yourself instead of pattern-matching to a given explanation.
The prompt
Ready to copy — highlighted parts are example details you can swap.
Run a Socratic tutoring session with me on the topic below. Your only allowed move in this entire conversation is to ask me a question — you are never permitted to state a fact, give an explanation, or confirm/deny an answer directly, even when I ask you to. TOPIC Why supply and demand curves shift versus move along the curve MY CURRENT UNDERSTANDING (STARTING POINT) I know the basic shapes of the curves and that price is on the y-axis, but I keep mixing up 'shift' and 'movement along' GOAL OF THIS SESSION Be able to correctly identify, for five different scenarios, whether it's a shift or a movement along the curve HARD RULE Every single response from you must be a question, and nothing else — no statements before the question, no "That's correct, and now...", no explanations tacked onto the end. If I answer wrong, do not tell me I'm wrong; ask a question that would lead me to notice the problem myself (e.g., ask me to apply my answer to a case where it clearly fails, or ask me to restate my reasoning out loud). If I get stuck and can't answer at all, do not give me the answer — ask a smaller, more specific question that breaks the current question into an easier first step, working backward toward something I can actually answer from I know the basic shapes of the curves and that price is on the y-axis, but I keep mixing up 'shift' and 'movement along'. If I directly ask you to just tell me the answer, respond only with a question that reframes the ask (e.g., "What do you think would happen if...") — never break character to explain why you're doing this. Start the session now with a single opening question based on Be able to correctly identify, for five different scenarios, whether it's a shift or a movement along the curve and I know the basic shapes of the curves and that price is on the y-axis, but I keep mixing up 'shift' and 'movement along' — pick a question that's answerable from where I'm currently starting, not one that assumes knowledge beyond I know the basic shapes of the curves and that price is on the y-axis, but I keep mixing up 'shift' and 'movement along'. OUTPUT FORMAT Just the single opening question. Every message after this, in the whole session, follows the same hard rule: one question, nothing else.
Customize
Optional — swap in your own details for the highlighted parts above.
Why this works
GPT-5.1's default conversational instinct, even when explicitly asked to be Socratic, is to slip an explanation or confirmation into the response alongside a question — a phrase like "That's close, remember that a shift means..." followed by a question still does the cognitive work for the student, because the student can extract the fact from the sentence before the question mark ever registers. The hard, absolute rule that every single message must be a question and literally nothing else closes that loophole mechanically rather than relying on the model's general sense of what "Socratic" means, which is exactly the kind of instruction that needs to be stated as an inviolable constraint rather than a stylistic preference, since a partial adherence still lets the answer leak through. The instruction for what to do when the student is simply wrong — ask a question that surfaces the failure themselves, such as applying their own answer to a case where it breaks — is the actual mechanism of Socratic method: the student discovers the contradiction in their own reasoning rather than being told there is one, which produces a correction that's self-generated and therefore stickier than one handed to them. The fallback for a genuinely stuck student, decomposing into a smaller answerable question rather than defaulting to an explanation, keeps the entire session inside the question-only constraint even at the point where a normal tutor would be most tempted to just explain — this is the hardest part of the rule to hold and the part most likely to break down without an explicit instruction for exactly that failure case. The instruction to resist a direct request for the answer by reframing rather than breaking character prevents the single most common way this kind of prompt collapses in practice, since students under time pressure will predictably ask for the answer directly at some point in the session.
What you get back
Opening question: If the price of gasoline suddenly drops for reasons unrelated to the demand for cars, what do you think happens to how many cars people want to buy at each possible price — and would you draw that as the same demand curve, or a new one?
Verified against
ChatGPT GPT-5.1 · 2026-08-12
Changelog
- 2026-08-12 — Initial publish, verified against ChatGPT GPT-5.1.
Need this built into your business?
If a prompt isn't enough — what Scult builds, built and maintained for you — that's Scult's day job.
EXPLORE WHAT SCULT BUILDS
