What Makes a Good Prompt
The principles behind prompts that actually work.
What Makes a Good Prompt
A prompt is your instruction to the AI. Think of it like giving directions to a stranger in a new city -- the more precise and clear your directions, the faster they get where you want them to go. The difference between a mediocre and exceptional AI user almost always comes down to prompt quality.
Watch: 3Blue1Brown — Attention in transformers (understanding why specificity matters)
The single biggest lever you have over AI output quality is your prompt. Better prompts do not require technical skills -- they require clear thinking.
The Five Principles
1. Clarity
Say exactly what you mean. AI cannot read between the lines, pick up on hints, or infer what you "really meant." If your prompt could be interpreted two different ways, the AI will pick whichever interpretation you did not want.
Bad: "Help with my project"
Good: "Help me outline a 5-page research paper on renewable energy adoption in Southeast Asia"
The bad prompt could mean literally anything -- a school essay, a home renovation, a software project. The good prompt tells the AI exactly what it is working on, how long it should be, and what the topic is.
2. Context
Give the AI the background it needs. You have all the context about your situation in your head. The AI has none of it. Bridge that gap.
Bad: "Write a response to this complaint"
Good: "I run a small online clothing store. A customer emailed saying their order arrived 5 days late and the wrong size. Write an apologetic response that offers a full refund and 20% discount on their next order."
Here is another context-rich prompt you can adapt:
"I am a freelance graphic designer pitching to a mid-size tech company. They asked me to send a brief capabilities overview. Write a one-page pitch that highlights brand identity, UI design, and packaging design experience. Tone should be confident but not salesy."
3. Specificity
The more specific your prompt, the more useful the output. Vague prompts produce vague results -- every time.
Bad: "Make it better"
Good: "Rewrite this paragraph to be more concise (under 50 words), use active voice, and make the tone more authoritative"
Notice how the good prompt gives the AI three concrete targets: a word count, a voice style, and a tone. That turns a subjective task ("make it better") into an objective one the AI can actually deliver on.
4. Constraints
This is counterintuitive, but boundaries actually help AI produce better output. Without constraints, the AI has to guess at your preferences for length, tone, format, and scope. With constraints, it can focus its energy on the content itself.
Examples of useful constraints:
- Word/character limits
- Tone (formal, casual, technical)
- Format (bullets, numbered list, table, paragraphs)
- Audience (beginners, experts, children, executives)
- What to include or exclude
A simple constraint like "keep it under 200 words" or "format as a table" can transform a rambling AI response into something you can actually use immediately. When in doubt, add a constraint.
5. Examples
Showing the AI what you want is often more effective than telling it. When you provide an example, the AI pattern-matches against it and reproduces the structure, tone, and format automatically.
"Format the output like this example:
Product: Widget Pro
Price: $29.99
Key Feature: Automated reporting"
This is especially powerful for tasks where describing the format in words would take longer than just showing it.
The Prompt Quality Formula
Output Quality = (Clarity + Context + Specificity + Constraints) x Iteration
Do not give up after one prompt. The biggest mistake new AI users make is treating it as a one-shot tool. The real power comes from iterating -- send a prompt, review the output, then refine with a follow-up like "Good, but make it shorter and add specific numbers."
Your first prompt rarely produces the perfect output. But a good first prompt gets you 80% there, and a quick follow-up gets you to 95%.
Right now, think of a task you did at work this week. Write a prompt for it using all five principles. Then compare it to how you would have asked a coworker to do the same thing. You will likely find the prompt version is more precise -- and that precision is exactly what makes AI useful.
Exercises
0/3Which principle states that giving AI boundaries actually improves output?
Take this bad prompt and rewrite it using all 5 principles: "Write something about marketing." Your improved prompt should specify: the format, audience, topic focus, length, and tone.
Hint: Example start: "Write a 500-word blog post targeted at small business owners about..."
Write the worst possible prompt you can think of, then rewrite it as the best possible prompt for the same task.
Hint: Make the bad one vague, ambiguous, and context-free. Make the good one specific, contextual, and constrained.