Constraints That Actually Work

The Power of No

Most prompting advice focuses on what to ask for. But telling the AI what to avoid is equally powerful — and often the fastest way to eliminate the specific problems you keep running into. Constraints are guardrails. They prevent the AI from defaulting to the habits that make its output less useful for your needs.

The good news: "avoid," "do not," and "never" all work reliably. The AI is quite good at following negative instructions when they are stated clearly.

Cutting AI Filler Phrases

Left to its own devices, the AI tends to open responses with certain stock phrases that add length without adding value:

• "Certainly! I'd be happy to help with that."
• "Great question! Here's what you need to know..."
• "Absolutely, here's a comprehensive overview..."

If these bother you — and for professional use they usually should — a single constraint eliminates them:

"Do not start with a preamble or affirmation. Get straight to the output."

Similarly, the AI often adds a closing sentence summarising what it just wrote. If you do not need that, add: "No closing summary."

Tone Constraints

When you want a specific tone, positive instructions help (e.g., "write in a warm, direct style"). But negative constraints add precision:

• "Do not use corporate buzzwords like 'synergy,' 'leverage,' or 'game-changing'."
• "Avoid hedging language — do not say 'it seems,' 'it could be argued,' or 'one might consider'."
• "Do not be overly enthusiastic. Keep the tone measured and professional."
• "Avoid passive voice where possible."

Tone constraints are especially useful when you have a brand voice or personal writing style you want AI output to match. List the specific patterns you want eliminated.

Content Constraints

Sometimes you know exactly what you do not want included:

• "Do not mention competitors by name."
• "Do not include pricing — that will be added later."
• "Avoid technical jargon — the reader has no coding background."
• "Do not give legal advice or include disclaimers."
• "Focus only on the benefits, not the features."

These constraints prevent the AI from making sensible-but-wrong choices about what to include. Without them, it fills in the gaps with what it assumes is typical for this kind of content.

Format Constraints

You can also constrain format in the negative, which is sometimes clearer than specifying the positive:

• "No bullet points — write in continuous prose."
• "No markdown — plain text only."
• "Do not use headers. Write it as a single flowing email."
• "No numbered lists."

A Default Constraint Set Worth Memorising

For most professional writing tasks, these four constraints improve output almost universally:

1. "No filler phrases — start directly."
2. "No closing summary."
3. "Avoid hedging language."
4. "No buzzwords."

You do not need all four every time. But knowing them means you can drop them in whenever you notice the AI drifting toward generic, padded output.

Constraints vs. Instructions: Working Together

Constraints work best in combination with positive instructions, not instead of them. "Write a product description. Avoid jargon and keep it under 100 words" is stronger than just "Write a product description" or just "Avoid jargon." Positive instructions shape what the AI produces. Constraints sharpen it. Both are needed for consistent results.