Here is the most common complaint I hear from people trying AI for the first time: "the answers are so generic." And here is the honest response: generic questions get generic answers. The good news is that writing a better prompt is not a technical skill. It is the same skill as giving clear instructions to a helpful human, and you already have it. This guide shows you how to apply it to Claude, with real before and after examples.
If you have never used Claude before, my beginner's guide to Claude covers the basics. Everything below works in a free account.
Why do vague prompts produce disappointing answers?
When you ask a person for help, they fill in the blanks using what they know about you. Your coworker knows your boss hates long emails. Your friend knows you are a vegetarian. Claude knows none of this unless you say it, so when you leave blanks, it fills them with the most average possible assumptions. Average assumptions produce average output.
That means the fix is not cleverness, it is specificity. Every piece of context you add removes a guess Claude would otherwise have to make. You do not need to be exhaustive. You need to cover the handful of things that matter: audience, format, length, tone, and anything off-limits.
What are the four habits of a good prompt?
Nearly every prompting tip you will ever read collapses into these four habits:
- Name the audience. "For my landlord," "for a nervous first-time client," "for my 8-year-old." The same information gets written completely differently depending on who reads it.
- Describe the format and length. "Three bullet points," "a one-paragraph email," "a table comparing the options," "under 100 words." Without this, you get Claude's default, which is often longer than you want.
- Show an example of what good looks like. Paste in something you liked, yours or anyone's, and say "match this style." This is the single highest-impact habit on the list.
- Say what to avoid. "No corporate buzzwords." "Do not apologize on my behalf." "Skip the introduction and get straight to the list." Naming the failure you expect prevents it.
What do before and after prompts actually look like?
Let me show you the difference with three everyday examples.
Example 1: The awkward email.
Before: "Write an email asking for a refund."
After: "Write a short, polite but firm email to an online furniture store. They delivered a damaged bookshelf two weeks ago and their support has not replied to my first message. I want a full refund, not a replacement. Keep it under 120 words, no threats, and mention that I have photos of the damage."
The first prompt gets a generic template. The second gets an email you could send with two small edits. Notice that nothing in the after version is clever. It is just the facts a human assistant would have needed too.
Example 2: The dinner plan.
Before: "Give me meal ideas."
After: "Plan five weeknight dinners for a family of four. One kid will not eat anything spicy, we want at least two vegetarian nights, and nothing that takes more than 30 minutes. Give me the list first, then a combined grocery list organized by store section."
Example 3: The explanation.
Before: "Explain how mortgages work."
After: "Explain how a fixed-rate mortgage works to someone buying their first home in the US. Assume I know nothing about finance. Use a running example with simple round numbers, define every term the first time you use it, and end with the three questions I should ask a lender."
How do you fix an answer that misses the mark?
This is where beginners often give up, and it is exactly the wrong moment to quit. Claude is conversational. You do not need to engineer a perfect prompt on attempt one, because the second message is where the real magic happens. Just say what is wrong, plainly:
- "Shorter. Cut it in half."
- "Warmer, this sounds like a lawyer wrote it."
- "You used the word 'delve' four times, please stop."
- "Good structure, but rewrite the second paragraph, it is too apologetic."
- "Give me three different versions of the opening line."
Think of the first answer as a first draft from an eager assistant. Revising is normal and fast. A prompt plus two corrections usually beats ten minutes spent polishing a single perfect prompt.
When should you give Claude an example to copy?
Whenever style matters, which is more often than you would think. Descriptions of style are fuzzy: my "friendly" and your "friendly" are different. An example is exact. Some easy wins:
- Paste a past email you were happy with and say, "write the new one in this same voice."
- Paste a newsletter you admire and say, "structure my announcement like this."
- Paste your own rough paragraph and say, "keep my wording where possible, just fix the flow."
One caution: use examples you have the right to use, and remember the example sets the pattern, so pick one you actually want repeated.
How can you save your best prompts instead of retyping them?
Once you find phrasing that works, do not let it evaporate. Keep a simple note on your phone or computer with your favorite prompts, the way cooks keep recipe cards. Better yet, if a set of instructions applies to a whole area of your life, put it into a Claude Project so it applies automatically. I wrote a full walkthrough in my guide to using Claude Projects to organize your work, and it pairs perfectly with the habits in this post: the four habits improve individual requests, and Projects stop you from repeating the background context at all.
Where can you practice without pressure?
Prompting is a skill you build by doing, not reading. Pick one low-stakes task this week, a grocery plan, a tricky text message, a summary of a long article, and run it through the four habits. Then correct the result twice. That one exercise will teach you more than any list of tips. When you are ready to go deeper, my free prompt engineering guide for beginners walks through the fundamentals step by step, still in plain English, still no jargon.
Next step: Put your new prompting habits to work across more tools by exploring my AI tools guides, all written for beginners in plain English.
