Starter

The AI Prompt Playbook: A Formula That Works

6 minute readUpdated June 2026Explore more

TL;DR

Most weak AI answers come from weak prompts, not weak models. There is a simple formula - role, task, context, format, constraints - that turns vague requests into sharp ones. Learn the five parts, avoid the four common mistakes, and reuse the templates below and your results jump immediately.

Why your prompts are not working

When an AI gives you a vague, generic, or off-target answer, the instinct is to blame the model. Almost always the real problem is the prompt: it did not say who the AI should be, what exactly you want, what it needs to know, or what good looks like. The model filled the gaps with guesses. Close the gaps and the same model gives you dramatically better work.

The five parts of a strong prompt

  • Role - tell the AI what perspective to take. 'Act as a hiring manager reviewing this resume.' This shapes tone and priorities in one line.
  • Task - say exactly what you want done, as a clear verb. 'Rewrite', 'summarize', 'find the three weakest points', 'draft'. Vague tasks get vague answers.
  • Context - give it the facts it cannot guess: who it is for, what you have tried, the background. This is where most prompts are too thin.
  • Format - describe the shape of the output you want. 'A five-bullet list.' 'A table with two columns.' 'Under 150 words.' If you do not say, you get whatever it defaults to.
  • Constraints - the rules and limits. 'No jargon.' 'Keep it to one paragraph.' 'Only use the information I gave you.' Constraints are how you steer quality.

The four mistakes that waste your time

  1. 1Being polite instead of clear. 'Could you maybe help me with something about my email?' wastes words and says nothing. State the task directly.
  2. 2Leaving out the context. The AI does not know your situation unless you tell it. A prompt with no background is a prompt that guesses.
  3. 3Asking for everything at once. A single prompt that wants research, a draft, and edits all together produces mush. Break it into steps and build on each answer.
  4. 4Not saying what good looks like. If you do not describe the format and constraints, you cannot be surprised when the shape is wrong. Show it the target.

Copy-paste templates

Fill in the brackets. These cover the requests people make most often - adapt them, then save the ones that work for you.

  • Rewrite: 'Act as a [role]. Rewrite the text below to be [goal - clearer / shorter / friendlier]. Keep it under [length]. Do not add new facts. Text: [paste].'
  • Explain: 'Explain [topic] to me like I am completely new to it. Use a plain-language analogy, then give one concrete example. Keep it under 200 words.'
  • Plan: 'I want to [goal]. My situation: [context]. Give me a step-by-step plan with no more than [number] steps, ordered by what to do first.'
  • Critique: 'Act as a tough but fair [role]. Here is my [thing]. Find the three weakest points and, for each, tell me exactly how to fix it. Be specific.'
  • Decide: 'Help me choose between [option A] and [option B] for [goal]. List the real tradeoff, then give me your single recommendation and why.'

Common questions

  • Do I need all five parts in every prompt?

    No. A quick question does not need a role or format. But when an answer comes back weak, the fix is almost always adding a part you left out - usually context or format. The formula is a checklist for when results disappoint, not a rule for every message.

  • Does saying 'please' really waste tokens and hurt results?

    It does not hurt results, and the token cost of a 'please' is trivial. The real issue is when politeness replaces clarity - 'could you maybe help with something' instead of a direct task. Be as polite as you like, but always state the task plainly.

  • What is the single highest-impact thing to add to a weak prompt?

    Context. Most prompts are too thin on the facts the AI cannot guess - who it is for, what you have already tried, the background. Adding two or three sentences of real context usually improves the answer more than any other single change.

  • Should I write one perfect prompt or keep going back and forth?

    Go back and forth. Treat it as a conversation: send a solid first prompt, read the result, then refine with follow-ups like 'shorter' or 'less formal.' Iterating beats trying to engineer one flawless prompt up front.

Get the prompt formula plus all the templates

Get 650+ plug-and-play skills, MCPs & prompts, plus 5,000+ builders - $9/mo, cancel anytime.

Join the Club