Using constraints effectively

Why constraints help

A constraint is a rule the output must follow: a word limit, a reading level, a required ingredient, a topic to avoid, a budget. Constraints narrow down the huge space of possible answers to the ones that are actually useful to you. "Write a dinner recipe" has millions of valid answers. "Write a vegetarian dinner recipe for two, under 30 minutes, using only one pan" has far fewer — and they're much more likely to be what you need.

Constraints also help with quality. Many people find that tight limits ("explain it in three sentences," "no jargon") produce clearer writing than open-ended requests.

Types of constraints

Write constraints the AI can follow

Make them measurable. "Keep it short" is a wish; "under 80 words" is a constraint. "Simple language" is vague; "avoid technical terms, or define any you use in a few words" is checkable.

Prefer positive phrasing. Telling the AI what to do is usually more reliable than listing what not to do. "Write in plain, conversational English" works better than "Don't be too formal, don't use jargon, don't use long sentences." Negative rules are still useful for hard boundaries ("don't include prices"), but pair them with what you want instead.

Explain the reason when it isn't obvious. "Keep it under 160 characters — it's going in an SMS" helps the AI prioritise. A constraint with a reason is easier to apply sensibly in edge cases.

Grounding constraints: "use only what I gave you"

One of the most useful constraints is limiting the AI to the material you provide. This reduces — though it doesn't eliminate — the chance of invented details.

Answer the question using only the policy text below. If the answer isn't in the text, say "The policy doesn't say" — don't guess.

Policy:
"""
[paste HR policy]
"""

Question: How many days of parental leave do part-time employees get?

That illustrative response shows what you want: a clear "not stated" instead of a plausible-sounding invention. Giving the AI an explicit way out makes it more likely to take it.

Avoid conflicting constraints

Constraints can clash: "Explain quantum computing thoroughly, covering all major concepts, in under 50 words." The AI can't satisfy both, so it will quietly sacrifice one — and you won't control which. Before sending, read your constraints together and ask whether a single answer could meet all of them. If two might conflict, say which wins: "If you can't fit everything in 100 words, prioritise the safety steps."

Don't over-constrain

Every constraint removes options. Pile on too many and the output gets stiff, awkward or worse at the core task — for example a story forced to use ten specific words, rhyme, stay under 50 words and teach a lesson. Start with the constraints that matter most, check the output, and add more only if needed.

Putting it together

A strong constrained prompt usually has: the task, the key context, a short "Constraints" list (measurable, mostly positive, no conflicts), and a priority order if anything might clash.