The Problem With "Better" Prompts
Most prompt advice tells you to be detailed. Write longer prompts. Add more adjectives. But this misses the real issue: most people don't know what to detail. They end up with bloated prompts that confuse AI tools instead of guiding them.
The best prompts aren't longer—they're more constrained. Constraints force both you and the AI to think clearly about what you actually want.
What Constraint-Based Prompting Actually Means
A constraint is a specific rule or boundary you set before asking the AI to create. Instead of hoping the tool guesses your taste, you decide the rules of the game first.
For text generation, constraints might be:
- Word count (exactly 150 words, not "brief")
- Tone (like explaining to a 10-year-old, not "simple")
- Structure (three paragraphs with a numbered list in the middle)
- Exclusions (no clichés, no passive voice, no emojis)
For image generation, constraints might be:
- Art style (oil painting, not "artistic")
- Composition (rule of thirds, left-aligned subject)
- Color palette (warm earth tones only)
- Aspect ratio (16:9, square, portrait)
Why Constraints Work Better Than Descriptions
When you say "write something inspiring," the AI has thousands of paths forward. When you say "write exactly 200 words in the voice of a cynical startup founder, with three short paragraphs, no exclamation marks," the AI knows exactly what game to play.
Constraints also reveal when you're unclear. If you struggle to write the constraint, you probably don't know what you want yet. That's valuable self-awareness before you waste time regenerating outputs.
Practical Constraint Patterns
For ChatGPT and Text Tools
The Format Constraint: Instead of "write a job description," try:
- "Write a job description using this exact template: [Title]. [Company mission in one sentence]. [Role overview in 2-3 sentences]. [5 bullet-point requirements]. [5 bullet-point nice-to-haves]. [Compensation range]."
Now the AI can't ramble or miss sections.
The Voice Constraint: Rather than "sound professional," specify:
- "Write as if you're a burned-out manager writing to your direct report. Include at least one slightly sarcastic comment. Keep it under 200 words. Use contractions."
The sarcasm and contraction rules lock in tone instantly.
The Exclusion Constraint: Add what you don't want:
- "Write a product description for headphones. Do not use: 'premium,' 'state-of-the-art,' 'immersive,' or any other marketing clichés. Focus on what the headphones actually do."
Negative constraints often produce clearer output than positive ones.
For Midjourney and Image Tools
The Style Stack: List exact styles or artists, separated by commas:
- "A kitchen counter with fresh fruit. Rendered in the style of David Hockney and Wes Anderson, warm color palette, high contrast, shot from above, 35mm film photography."
Stacking styles gives the AI specific visual reference points rather than abstract goals.
The Composition Constraint: Name the specific framing rule:
- "A person looking at a laptop. Centered in the frame. Shot from directly above. Minimal background. Flat lay style."
This removes ambiguity about camera angle, which is often the biggest variable.
The Technical Constraint: Add render specs:
- "A forest landscape. Cinematic lighting. Shot at golden hour. 8k resolution. Shallow depth of field with blurred background. No people, no text, no watermarks."
Specifying resolution, lighting time, and effects prevents common unwanted outcomes.
Building Your Constraint Checklist
Before you write any prompt, answer these questions:
- Length or scope: How much output? (word count, image count, number of steps)
- Format: What structure or template?
- Tone or style: Who's the voice? What's the visual reference?
- Do not include: What specific things ruin this for you?
- Context: Who is the audience? Where will this be used?
- Success metric: How will you know it's good?
Write these down before you prompt. This five-minute setup saves 15 minutes of regenerations.
Real Example: Job Resume
Weak prompt: "Write a resume for a marketing manager."
Constrained prompt: "Write a resume for a marketing manager with 6 years of B2B SaaS experience. Use a chronological format. Include exactly 4 bullet points per role. Each bullet must start with a metric (numbers, percentages, dollar amounts). No buzzwords like 'synergy,' 'leverage,' or 'dynamic.' Total length: one page, 400-450 words. Include a 2-sentence professional summary at the top."
The second version leaves almost nothing to chance.
When Constraints Backfire
One caution: constraints work best when they're intentional, not arbitrary. If you add a constraint just to sound smart, it wastes time.
Also, some constraints conflict. Don't ask for a 500-word article and a 3-minute read—these are different constraints. Pick one.
Testing and Refining
After you get a result with constraints, note which ones actually improved the output and which were unnecessary. Over time, you'll build a personal library of constraints that work for your style and needs.
This is the real skill: not memorizing prompts, but knowing which constraints matter for your specific use case.
The Takeaway
Prompt engineering isn't about writing longer, flowery descriptions. It's about thinking clearly enough to set boundaries. The best prompts are often the most specific ones—they remove options rather than add them.
Once you start writing constraints before you prompt, you'll notice your results improve immediately. And you'll spend less time regenerating, because you've already done the thinking work.
If you're ready to move beyond trial-and-error prompting, explore ready-to-use AI prompts on Nohaya PromptAI, where you can see how expert prompts use constraints to get consistent, professional results.