The AI Learning Hub Journal

The Mistake Almost Everyone Makes

The most common weak prompt is perfectly polite — and aimed at nobodytwo facts are missing: who it is for, and what it is for — so the model has to pick, and it picks the safe middleTHE REQUEST"Write a project update."FOR WHOM? — IT HAS TO GUESSyour manager? the client?the whole company? a board?each reader needs a different updateFOR WHAT? — IT HAS TO GUESSto reassure? to escalate a problem?to ask for budget? to create a record?each purpose needs a different updateTHE SAFE MIDDLE — WHAT GUESSING PRODUCESa neutral, medium-length, generic update aimed at nobody in particularpeople then call the output bland — it is bland because the request was aimed at no oneTHE THREE-LINE UPGRADE — ABOUT FORTY SECONDSAUDIENCE"for the client's operations lead — detail-oriented, dislikes vague language"PURPOSE"flag the two-week slip and get her to approve the revised date"QUALITY BAR"a good update names the slip in the first sentence — and skips background she already knows"ADD WHO IT IS FOR AND WHAT IT IS FOR — BEFORE YOU PRESS SENDaudience and purpose are not polish; they are the two facts that decide what a good answer looks like
Most weak prompts are missing who the answer is for and what it is for — add those two facts and the generic output disappears.

Asking for a Thing With No Reader and No Purpose

The most common bad prompt is not rude, lazy or badly spelled. It is a perfectly reasonable request with two things missing: who it is for, and what it is for. "Write a project update." For whom — your manager, the client, the whole company, a board? For what — to reassure them, to escalate a problem, to ask for more budget, to create a record? Without those, the model has to pick, and it picks the safe middle: a neutral, medium-length, generic update aimed at nobody in particular. Then people conclude the output is bland. It is bland because the request was aimed at no one. Audience and purpose are not polish. They are the two facts that determine what a good answer even looks like.

  • Weak: "Write a project update" — no reader, no goal, no stakes
  • Strong: "Write a project update for our client's operations lead. Purpose: flag a two-week slip and get her to approve a revised date. She is detail-oriented and dislikes vague language."
  • Missing audience produces generic output; missing purpose produces unfocused output
  • Most "AI writing sounds like AI" complaints trace back to exactly this gap

Why "Be More Specific" Is True and Useless

It is the advice everyone gives and almost nobody can act on, because it does not say specific about what. Adding more words is not the fix — a longer vague prompt is still vague. What actually helps is knowing which dimensions matter. Specific about the audience: who reads this and what do they already know? Specific about the purpose: what should happen after they read it? Specific about scope: what is deliberately out of scope? Specific about format: how long, what shape, what sections? Specific about quality: what would make this answer wrong or unusable? Naming those five turns an abstract instruction into a checklist you can actually run.

  • Audience: who reads it, what they already know, what they care about
  • Purpose: what you want to happen as a result
  • Scope: what to include and what to deliberately leave out
  • Format and quality bar: shape, length, and what would make it unusable

The Three-Line Upgrade

You can fix most weak prompts by adding three lines, and you can do it in under a minute. Line one: this is for [audience], who [what they know or want]. Line two: the purpose is [what should happen next]. Line three: a good answer would [test of quality] — and it should not [the thing you keep getting that you do not want]. Take "summarise this report". Add: "This is for a busy director who has not read the report and will act on the summary. Purpose: decide whether to fund phase two. A good summary states the recommendation in the first sentence and flags anything that would change the decision. Do not include background the director already knows."

  • Line 1 — audience: who reads this and what they already know
  • Line 2 — purpose: the decision or action this is meant to enable
  • Line 3 — quality bar plus one explicit exclusion
  • Three lines is roughly forty seconds and eliminates most rewriting

Try It Yourself

Reading about the under-specified one-liner is one thing; catching your own is another. This takes five minutes and one real request you actually sent.

◆ Try it yourself

Open your AI tool and find the last real request you sent it. Send it again in a fresh conversation, unchanged. Then send the upgraded version below — the same request plus three lines. Compare the two answers.

[your original request]

For: [who will read or use the answer]
So that: [what you will actually do with it]
Keep it: [length, tone, and format limits]
How you'll know it worked
  • The two answers are noticeably different
  • The upgraded answer needed less editing before you could use it
  • You can point to which added line changed the answer most

Prefer slides, quizzes, and saved progress? Read this lesson in the library — free, no sign-up.