Tell AI what not to do with AI prompt constraints
The fastest way to stop the habits you keep editing out of AI drafts by hand.
In short
AI prompt constraints are the rules that tell the model what to avoid: words, formats, and behaviors you do not want.
- Turn the things you always delete from AI drafts into standing 'do not' rules.
- Be specific. 'No corporate jargon' beats 'sound natural,' and naming the exact words beats both.
- Keep it to your top three to five. A long deny list confuses the model more than it helps.
Why constraints work
AI prompt constraints tell the model what to leave out, and they fix the exact things you keep correcting by hand. Positive instructions describe what to build. Constraints carve away what you do not want. Together they land far closer to right. If every AI draft opens with a windup you delete, one 'do not open with a preamble' rule saves you that edit every time. The constraints you use most belong in your standing Custom Instructions, so they run on every chat automatically. Constraints belong in your ChatGPT system prompt once they settle, and length is the constraint people reach for first, covered in controlling response length. The broader skill is prompt engineering.
How to set AI prompt constraints
Your own edits are the list. Write them as rules and keep it short.
- 1
List what you always fix
Look at your last few AI drafts. The edits you make every single time are your constraint list, already written for you.
- 2
Write them as clear 'do not' rules
Turn each edit into a rule: 'do not use bullet points,' 'do not add a conclusion,' 'keep it under 150 words.'
- 3
Be specific about what to avoid
'No jargon' is fuzzy. Name the exact words, like 'do not use synergy or circle back,' and the model can actually follow it.
- 4
Separate hard rules from preferences
Use 'do not' and 'never' for hard limits; use 'prefer to avoid' or 'minimize' for softer ones the model can break if it needs to.
- 5
Cap the list at three to five
Keep only the constraints that matter most. A long list of negatives gets half-followed and starts to confuse the model.
Constraints to reuse
Watch-outs
- Too many negatives confuse the model. Three to five sharp rules beat fifteen vague ones.
- Pair a 'do not' with a 'do' when you can: 'no jargon, use plain words a new hire would understand.'
- Vague negatives do nothing. 'Avoid bad writing' is not a rule; 'no sentences over 25 words' is.
Common questions
What are AI prompt constraints?
They are the rules in a prompt that tell the model what to avoid: specific words, formats, lengths, or behaviors. Also called negative prompting, they carve away the output you do not want, which pairs with your positive instructions to land closer to right.
How do I tell AI what not to do?
Write clear 'do not' rules for the things you keep deleting: 'do not add a conclusion,' 'do not use bullet points,' 'do not exceed 150 words.' Be specific, name exact words to avoid where you can, and keep the list short.
How many constraints should I use?
Three to five that matter most. A short, targeted deny list gets followed; a long one gets half-followed and can confuse the model into worse output. Prioritize the edits you make on every draft.
What is negative prompting?
Negative prompting is telling the AI what to exclude rather than only what to include. Hard negatives ('do not,' 'never') are non-negotiable; soft ones ('prefer not to,' 'minimize') are preferences the model can override when needed.
See how you actually work with AI
The AI Skills Quiz scores your real habits, including how tightly you steer the model's output. 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