Write a ChatGPT system prompt that makes it follow your rules
Set the model's role, rules, and format once so you stop correcting the same things in every chat.
In short
A ChatGPT system prompt is a saved instruction block that sets the model's role, rules, and output format for every reply.
- Put your standing rules in Settings, Custom Instructions, so they apply to every chat automatically.
- Structure it: role, goal, rules, format, and one example of a good answer.
- Tell it what not to do. Constraints stop the habits you keep fixing by hand.
- Refine it over time: when a reply misses, add the one rule that would have caught it.
What a system prompt is, and where it lives
A ChatGPT system prompt is the set of standing instructions the model reads before it answers: who it should act as, what you want, and the rules it must follow. In the API that block is literally called the system prompt. In the ChatGPT app, Custom Instructions do the same job. You set them once and they apply to every conversation. In July 2026 OpenAI raised that limit from 1,500 to 5,000 characters on its Pro, Business, Enterprise, and Education plans. That is room for a real briefing, not a couple of lines. You can also paste an instruction block at the top of a single chat or into a Project. Write a good one and you stop correcting the same things every day. When the model keeps ignoring your instructions, a sharper system prompt is usually the fix. The rules worth putting in one usually start as prompt constraints you keep retyping, and the prompts themselves belong in a personal prompt library. It is the highest-leverage habit in learning ChatGPT.
How to write a ChatGPT system prompt
Six lines do most of the work: a role, a goal, your rules, your anti-rules, a format, and one example.
- 1
Give it a role and a goal
Open with who the model should act as and what you are trying to achieve: 'You are my editor. Your goal is to make my writing clearer and shorter without changing my meaning.'
- 2
Write the rules as a short list
List the behaviors you want, one per line: keep answers under 200 words, use plain English, show your reasoning, ask before assuming.
- 3
Say what not to do
Add an anti-rule for each habit you keep correcting: do not open with 'Certainly,' do not add a summary at the end, do not hedge with 'it depends.'
- 4
Set the output format
Tell it the shape you want: three short paragraphs, or a table with named columns, or an answer that ends with one question to sharpen your request.
- 5
Show one example of a good answer
Paste a short reply you liked or a quick before-and-after. One example teaches tone faster than a paragraph describing it.
- 6
Save it, then refine it
Put the block in Settings, Custom Instructions so it applies everywhere, then improve it: every time a reply misses, add the single rule that would have caught it.
A system prompt template to fill in
Watch-outs
- Custom Instructions apply to every chat, so keep them general. Put task-specific rules at the top of that one chat.
- More is not better. A tight set of rules beats a long wish list the model only half-follows.
- Re-read it every few weeks. Rules you set months ago may be fighting how you work now.
Common questions
What is a ChatGPT system prompt?
It is the block of standing instructions the model reads before every answer: its role, your rules, and the output format you want. In the ChatGPT app you set it through Custom Instructions; in the API it is the system message. A good one means you stop repeating the same corrections in every chat.
Where do I put a system prompt in ChatGPT?
For standing rules, use Settings, Personalization, Custom Instructions, and they apply to every conversation. For rules that only matter this once, paste an instruction block at the top of the chat or set it inside a Project so it stays attached to that work.
What should a good system prompt include?
A role, a goal, a short list of rules, an anti-rule for each habit you keep correcting, the output format, and one example of a good answer. Keep it tight: a few clear rules work better than a long list the model half-follows.
Why does ChatGPT ignore my instructions?
Usually they are vague, buried, or contradict each other. Make each rule specific, put the most important ones first, and add an anti-rule for the exact habit you want gone. If a reply still misses, add the single rule that would have fixed it and save it.
See how you actually work with AI
The AI Skills Quiz scores your real habits, including how well you brief the model before you ask. Free, and it takes a couple of minutes.
Professionals from these companies build their AI skills with Candova
The future belongs to AI‑native companies.
Make sure yours is one.
We make your people AI-fluent so your company becomes AI-native, with Cando, a personal AI trainer for every employee. Start with Candova AI today.
Free to start · Productivity gain guarantee · Built for your real job