15 ChatGPT Response Codes

Each code is a shortcut for a full instruction. Paste the instructions once, then type the code with your question when you want that kind of answer.

What this helps you do

Keep an answer grounded in your files, expose hidden assumptions, collect missing context, challenge a plan, or compare options without rewriting the same instructions.

Each code is a short label with a full instruction after it, and each one changes the kind of answer you get. /NEXTSTEP is the one to reach for when ChatGPT hands you a plan that sounds smart but leaves you with nothing to do.

Before you start

Open ChatGPT and start a new chat. Upload any source files you want it to use. Remove passwords, private client data, medical information, and anything else that should not be sent to an online service.

Terms to know

• Code: a short label you set up once in your chat, like /CHALLENGE, that stands for a full instruction.
• Source lock: a rule that limits the answer to supplied material.
• Countercase: the strongest credible argument against a plan.
• Reversibility: how easily a decision can be undone.
• Teach-back: explaining an idea in your own words to prove you understand it.

Step-by-step

1. Paste the setup block below into a new chat.
2. Add your files or question.
3. Type one label, such as /SOURCELOCK, followed by the task.
4. Check the answer against your source before acting on it.
5. Keep the full instructions in a note on your phone, so you can paste them into any new chat in seconds.

Capital letters don't matter. /challenge works the same as /CHALLENGE, as long as the instruction is in the chat.

Copy this

Use these custom response labels in this conversation:

/SOURCELOCK: Answer only from the files I provided. Cite the file and section for every important claim. Label anything missing as unknown.

/ASSUMPTIONCHECK: List the assumptions behind your answer before you recommend anything. Mark each one as supported, uncertain, or unsupported.

/MISSINGINFO: Ask me up to three questions that would materially improve the answer. Wait for my replies before answering.

/COUNTERCASE: Give me the strongest evidence-based reason my current plan could fail. Do not invent objections just to disagree.

/DECISIONTABLE: Compare my options by cost, risk, time, and reversibility. State what information is missing and do not choose a winner unless the evidence supports one.

Follow each code only when I type it.

Five learning and planning codes

Paste this block into your chat first, then type any code with your question.

/TRUTHMODE: Stop agreeing with me just to be nice. If I am wrong, say so and show me why. Separate what my sources support, what you infer, and what you do not know. Cite the source for every important fact.
/8020: Find the smallest set of actions likely to produce most of the result. Explain why you chose them and what would change that choice.
/SOCRATES: Teach me by asking one question at a time. Wait for my answer before correcting me or moving on.
/UNLEARN: List the most common wrong assumptions about this topic. For each, show the accurate idea and one example.
/VISUALIZELEARNING: Make a simple labeled diagram or flowchart of this idea, then explain the parts in order.

Five codes for clearer answers

Use these when an answer feels vague. Paste the block, then type one label with your question.

Use these custom response labels in this conversation:

/CHALLENGE: Give me the strongest reason my plan could fail.

/COMPARE: Put these two options side by side, using cost, time, and what I give up.

/SHOWME: Give me one finished example before you explain the method.

/TEACHBACK: Ask me one question at a time until I can explain this myself.

/NEXTSTEP: Give me the smallest thing I can do in ten minutes.

Follow each code only when I type it.

For example, type /CHALLENGE I'm planning to [YOUR PLAN]. or /COMPARE [OPTION A] or [OPTION B].

Five answer-check codes

The first five codes above already include /SOURCELOCK, /ASSUMPTIONCHECK, /MISSINGINFO, /COUNTERCASE, and /DECISIONTABLE. Use them for the answer-check set. Keep the full instruction in the conversation so typing a label recalls the exact task you assigned it.

Things to know

These codes shape how ChatGPT answers. /SOURCELOCK works best with readable text files and exact citations. Scans, tables, and incomplete uploads can be misread. /MISSINGINFO should ask only questions that could change the answer.

Some codes overlap on purpose. /CHALLENGE is the quick version of /COUNTERCASE, and /COMPARE is the quick version of /DECISIONTABLE. Use the longer one when the decision is big.

If something goes wrong

• If ChatGPT ignores a label, paste the full instruction again with your question.
• If it cites text that is not in a file, ask for the exact quoted line and verify it yourself.
• If the decision table invents costs, supply real numbers or tell it to write unknown.
• If /NEXTSTEP gives you a list, ask for "one action only, and how I'll know it's done."
• If /TEACHBACK asks several questions at once, repeat "one question at a time."
• If a saved instruction affects unrelated chats, remove it and keep these labels inside one project or conversation.

Final check

Test /SOURCELOCK with a question your file does not answer. The setup passes only if ChatGPT says the answer is unknown instead of filling the gap.

Then type /NEXTSTEP after any vague plan. You pass when the answer is one action you can finish in ten minutes.