Most people think the best AI prompts are detailed and open-ended. But some of the most effective prompts work the opposite way: they add one tight constraint that forces the model to focus.
A constraint isn’t about limiting creativity—it’s about giving the AI a clear boundary that makes its output more precise, actionable, and relevant. Here are six common prompts that improve dramatically when you add a single, well-chosen constraint.
1. Summarize This Article (Add: Target Word Count)
Instead of asking ChatGPT or Claude to “summarize this article,” specify exactly how long the summary should be. Try “Summarize this in exactly 50 words” or “Give me a three-sentence summary.”
Why it works: Without a length constraint, AI models default to verbose summaries that often rehash every section. A word or sentence limit forces prioritization—the model has to identify what actually matters.
Best for: Email briefings, meeting prep, or sharing articles with colleagues who need the gist fast.
2. Write an Email (Add: Reading Level or Tone)
Generic email prompts produce generic emails. Instead, try “Write this email at an 8th-grade reading level” or “Write this email as if you’re a peer, not a manager.”
Why it works: Tone and complexity constraints help the model match your audience. An email to a busy executive needs different language than one to a technical team or a frustrated customer.
Best for: Cold outreach, internal updates, customer support, or any email where the wrong tone kills the message.
3. Brainstorm Ideas (Add: A Format Constraint)
Open-ended brainstorming prompts often produce a wall of bullet points. Add a format constraint: “Give me five ideas, each as a single tweet” or “Brainstorm three concepts, each described in one sentence with a concrete example.”
Why it works: Format constraints force the AI to think in specific structures. A tweet-length idea has to be punchy. An idea paired with an example has to be concrete, not abstract.
Best for: Content planning, product features, campaign angles, or any creative work where vague ideas don’t help.
4. Explain This Concept (Add: An Analogy Requirement)
Instead of “Explain what a transformer model is,” try “Explain what a transformer model is using only analogies from cooking.”
Why it works: Forcing the AI to use a specific analogy domain—cooking, sports, music—makes explanations more memorable and easier to grasp. It also prevents the model from falling back on jargon or overly technical language.
Best for: Teaching, presentations, or explaining technical ideas to non-technical teammates.
5. Rewrite This (Add: A Perspective Shift)
Rewriting prompts are too vague on their own. Add a perspective constraint: “Rewrite this from the customer’s point of view” or “Rewrite this as if you’re skeptical, not excited.”
Why it works: Perspective shifts surface blind spots. A feature announcement written from the customer’s POV often reveals that you’re talking about yourself, not solving their problem. A skeptical rewrite helps you stress-test weak arguments.
Best for: Marketing copy, proposals, pitch decks, or any writing where you need to see it through someone else’s eyes.
6. Generate Examples (Add: A Context Constraint)
Don’t just ask for examples—specify where they’ll be used. Instead of “Give me examples of good subject lines,” try “Give me five subject lines for a B2B SaaS company announcing a new integration to existing customers.”
Why it works: Context constraints prevent the AI from generating generic, SEO-blog-style examples. The more specific the constraint, the more useful the output.
Best for: Copywriting, training materials, case studies, or any situation where examples need to feel real, not theoretical.
Why Constraints Beat More Words
It’s tempting to think that longer, more detailed prompts always produce better results. But constraints work differently than instructions. Instructions tell the AI what to do. Constraints tell it what not to do—or what shape the output must take.
That difference matters. A good constraint eliminates the vague middle ground where AI outputs feel almost useful but not quite. It forces decisions. It creates focus.
The trick is choosing the right constraint. Word count works for summaries. Tone works for emails. Format works for brainstorming. Perspective works for rewriting. The best constraint depends on what you’re trying to get out of the model—and what usually goes wrong when you don’t add one.
Want one useful AI idea like this delivered daily? Subscribe to the One Two Three AI newsletter and get smarter about AI without the hype.