Most weak AI answers start with a weak prompt. If you ask a short, vague question, you usually get a short, vague answer back.
This prompt engineering cheat sheet gives you simple patterns for asking better. You will learn the building blocks of a clear prompt, a few easy frameworks, ways to refine an answer, and the common mistakes that hold beginners back. No coding or special tools are needed.
Use the cheat sheet below as a quick reference. Then read on to see when to use each pattern and how it looks filled in.
How to Use These Prompts
Words in [brackets] are blanks. Replace each one with your own details before you send the prompt.
Before: Act as a [role]
After: Act as a friendly librarian who helps first-time readers pick a book.
You can combine several patterns in one prompt, or use them one at a time. They work the same way in ChatGPT, Claude, and Gemini. Treat every answer as a draft, and check facts yourself.
Prompt Building Blocks
Act as a [role]
A role tells the AI which point of view and level of expertise to use. It helps most when the topic has a clear job behind it, such as editor, tutor, or recruiter. Pick a role that fits your task.
Example: Act as a friendly librarian who helps first-time readers pick a book.
Task: [what to do]
State the job in one clear sentence. Start with a verb like write, compare, summarize, or list. A clear task is the single biggest step toward a useful answer.
Example: Task: Write a short welcome message for new members of a community garden.
Context: [background]
The AI cannot guess your situation. Context gives it the facts it needs, like who you are, what happened, and what you have tried. Include only what matters, and leave out private details.
Example: Context: We are a volunteer group of 20 people. Our garden opens in April, and most members are beginners.
Format: [list/table/steps]
Tell the AI how you want the answer laid out. A table, a numbered list, and a short paragraph all serve different needs. Naming the format saves you from reformatting later.
Example: Format: a numbered list of five steps, each under 15 words.
Audience: [who reads it]
The same idea sounds different for a child, a manager, or an expert. Naming the audience changes the tone, vocabulary, and level of detail. Be specific about who will read the result.
Example: Audience: adults who have never grown vegetables before.
Frameworks to Try
Role + Task + Format
This three-part pattern is a good starting point for everyday requests. State who the AI should be, what it should do, and how to present it. It takes seconds and often improves answers right away.
Example: Act as a tutor. Explain how fractions work. Give three short examples in a bulleted list.
Context + Goal + Limits
Use this when the situation matters. Give the background, say what you want to achieve, and set limits such as length or tone. The limits keep the answer focused.
Example: Context: I run a small pet-sitting business. Goal: write a friendly reminder about payment. Limits: under 80 words, no pressure.
Show 2 examples, then ask
If you want a certain style, show it. Paste two short examples, then make your request. The AI will follow the pattern it sees, which is often easier than describing the style in words.
Example: Here are two product descriptions I like: "Soft cotton tote, big enough for groceries." and "Handmade mug, warm in your hands." Now write one for a canvas lunch bag.
Think step by step
For problems with several parts, ask the AI to show its steps. This makes it easier to see where an answer went wrong. Check the steps yourself, because a clear explanation can still contain mistakes.
Example: Think step by step: A class of 24 students splits into groups of 4. Each group needs 3 markers. How many markers are needed?
Ask what you need to know
Sometimes you do not know what details to provide. Let the AI ask. It will list the questions it needs answered before it gives a draft.
Example: I want to plan a small birthday party. Ask what you need to know before you make a plan.
Refine the Answer
Make it shorter, keep [points]
Long answers are common. This prompt trims the text while protecting the points you care about. Name those points so they do not get cut.
Example: Make it shorter, keep the date, the location, and the dress code.
Rewrite for a [level] reader
Use this to change difficulty up or down. You can ask for a beginner, a teenager, or an expert. It helps when a first answer feels too simple or too technical.
Example: Rewrite for a beginner reader: Compound interest is interest calculated on both the initial principal and accumulated interest.
Give 3 options, then compare
One answer can hide better choices. Asking for three options with a short comparison shows the trade-offs. Then you choose the one that fits.
Example: Give 3 options for a name for my dog-walking service, then compare them in a short table.
What’s missing from this?
Before you rely on an answer, ask what it leaves out. The AI may point out missing steps, risks, or questions. Treat the list as a prompt for your own thinking.
Example: What's missing from this plan for a neighborhood bake sale?
Check this against my notes
If you have source material, ask the AI to compare. It can flag claims that do not match what you wrote. You still need to read both and decide.
Example: Check this against my notes and list anything in the summary that my notes do not support.
Common Mistakes
Vague requests
“Help me with my email” gives the AI almost nothing. Say who the email is for, what you want, and how long it should be. Quick fix: add a role, a task, and a format.
Too much in one prompt
Packing five jobs into one request leads to mixed results. Split big jobs into steps and review each one. For example, plan first, then draft, then edit.
No examples
Describing a style in words can be hard. Show one sample of what you want. A single example often does more than a paragraph of instructions.
Accepting the first answer
The first answer is a draft. Ask for edits, such as “make it warmer” or “add a short example.” The second version is often better.
Trusting facts blindly
AI can state wrong names, dates, and numbers with confidence. Check anything important in a reliable source. This matters most for facts you plan to share or publish.
Pro Tip
Save the prompts that work in a doc. Reuse and tweak them instead of starting from scratch each time, and you will build your own set of favorites.
Example: Act as an editor. Task: tighten this paragraph. Format: the revised version, then a one-line note on what changed. Audience: busy customers.
FAQ
What is prompt engineering?
It is the practice of writing clear instructions for an AI tool so it gives more useful answers. For everyday users, it means giving context, stating the task, and asking for a format.
Do I need to learn special terms or code?
No. Plain, specific sentences are enough for most tasks. The patterns here are just ways to remember what to include.
Do these patterns work in every AI chatbot?
They are general habits, so they work in most chatbots, including ChatGPT, Claude, and Gemini. Results will still vary, so check each answer and adjust your prompt.
Final Thoughts
A good prompt is mostly clear thinking: who it is for, what you want, and what a good answer looks like. Start with the building blocks, try a framework, and refine from there. Check facts, and keep private information out of the chat.
If you want more ready-to-use prompts, get the full prompt pack here: Get the prompt pack here