AI Guides › Claude Mastery

Why Vague Requests Get Vague Answers

By Nigel Guy · 1 min read

The common response to a disappointing answer is to add more words to the next attempt — more adjectives, more insistence, more "please make it really good this time." That almost never fixes it, because the problem usually isn't effort or enthusiasm. It's shape.

The mechanism

1. Name the actual decision or use the answer is for, not just the topic. "Write about productivity" has no target. "Write three bullet points I can put in a Monday standup" does.
2. Give one concrete constraint, even an arbitrary one. A length, an audience, a format. A constraint gives the model something to satisfy instead of guessing at your unstated taste.
3. State what "wrong" would look like. Sometimes it's faster to say what to avoid than what to include — "don't make this sound like a sales pitch" narrows the space fast.
4. Add detail only where it's actually load-bearing. More words that don't add a real constraint just add more for the model to weigh, without making the target any clearer.

What to skip

Skip padding a request with enthusiasm or superlatives — "please make this really great" adds nothing a model can act on. And skip assuming a vague request will improve with a longer, equally vague follow-up; add a constraint instead of adding words.

Guardrails

• A specific request narrows the target; it doesn't guarantee the answer lands there. Still review what comes back.
• Over-constraining has its own failure mode — twenty constraints can bury the two that matter most. Prioritise rather than listing everything you can think of.
• If you're not sure what the actual constraint should be, that's worth figuring out first — a request can't be more specific than your own thinking about what you actually want.

All 751 AI guides · JulieMango plans from £17/mo