Here’s the open secret of AI tools: the quality of the answer depends more on your prompt than on which chatbot you use. Two people can ask the “same” question and get wildly different results — one gets generic fluff, the other gets exactly what they needed. The difference is the prompt.
Prompt engineering sounds technical. It isn’t. It’s just the skill of asking well — and it’s learnable in an afternoon. Here are the seven techniques that matter most, each with a real before-and-after example you can steal.
1. Be specific — vague in, vague out
AI fills gaps with guesses. The vaguer your prompt, the more generic the answer. Specificity is the single highest-leverage change you can make.
Weak prompt:
Write about electric bikes.
You’ll get a generic overview that could have been written by anyone, for anyone.
Strong prompt:
Write a 300-word comparison of electric bikes vs regular bikes for commuting in Amsterdam, aimed at someone who cycles 8 km to work daily. Cover cost, effort, and parking. Tone: practical, no hype.
Now the AI knows the audience, the length, the topics, the tone, and the city. The answer will be specific because the question was.
The rule: if your prompt could apply to ten different situations, it’s too vague. Add the details that make it yours.
2. Give context — the AI doesn’t know your situation
Chatbots don’t know who you are, what you know, or what you’re trying to achieve. Every relevant detail you add sharpens the answer.
Weak prompt:
Help me write a complaint email.
Strong prompt:
Help me write a complaint email. Context: my internet provider in Berlin charged me €45 for a technician visit I never requested. I want a refund and I want it in writing. I’m a customer of 3 years and I’ve never missed a payment. Tone: firm but polite. Keep it under 200 words.
See the difference? The second version gives the AI everything it needs: facts, goal, tone, length. The email it writes will need light editing, not a rewrite.
The rule: include the background a smart human friend would need to help you. Who, what, where, and what outcome you want.
3. Assign a role — borrow expertise on demand
Telling the AI who to “be” changes how it thinks. A role sets the knowledge level, vocabulary, and perspective of the answer.
Weak prompt:
Explain how mortgages work.
Strong prompt:
Act as a patient financial adviser explaining to a first-time buyer in the UK. Explain how fixed-rate mortgages work, what “LTV” means, and the three most common mistakes first-time buyers make. Use plain English, no jargon without explanation.
The rule: one clear role beats none. Don’t stack five roles — “act as a marketer, designer, lawyer, and poet” just confuses things.
4. Iterate — treat the first answer as a draft
Beginners treat AI like a vending machine: one prompt in, one answer out, done. Experienced users treat it like a conversation. The first answer is rarely the best answer — it’s a starting point you refine.
Try follow-ups like:
That’s a good start. Make it shorter — half the length.
This part about pricing is unclear. Explain it with a concrete example.
I like option 2. Give me three more variations in that style.
This sounds too formal. Rewrite it like I’m explaining it to a friend.
Each round gets you closer. Three quick iterations usually beat one “perfect” prompt, because you discover what you actually want by seeing what you don’t.
The rule: never accept the first draft without at least one round of feedback.
5. Give examples — show, don’t just tell
If you want a particular style, format, or level of quality, show the AI an example. This is the fastest way to get consistent results, and it’s how professionals get AI to match a brand voice.
Weak prompt:
Write product descriptions for my candles.
Strong prompt:
Write product descriptions for my candle shop. Here’s an example of the style I want:
>
“Rainy Sunday — Smells like wet pavement, old books, and staying in. Notes of cedar, rain lily, and a whisper of smoke. For people who cancel plans happily.”
>
Now write three more in the same style for: (1) a citrus morning candle, (2) a cozy winter candle, (3) a fresh linen candle. Keep each under 60 words.
The example does more work than ten adjectives. The AI now understands the voice — playful, sensory, slightly cheeky — and can reproduce it.
6. Set format and length — or get a wall of text
Left to itself, AI defaults to medium-length paragraphs with a cheerful introduction. If you want something else, say so explicitly.
Weak prompt:
Summarize this article about remote work trends.
Strong prompt:
Summarize this article about remote work trends as:
– 5 bullet points, each under 20 words
– Then one sentence: the single most surprising finding
– No introduction, just the list
You can request tables, numbered steps, pros-and-cons lists, scripts, outlines — any structure — and set length precisely: “in 100 words,” “a two-sentence summary,” “a detailed 800-word guide.”
The rule: if you have a format in mind, name it. The AI can’t read your mind, but it’s excellent at following a template.
7. One task per prompt — don’t bundle
Long prompts with five requests get you five mediocre answers. Split them up and each one gets full attention.
Weak prompt:
Write a blog post about sourdough, then summarize it for Twitter, translate it to French, suggest 5 headlines, and tell me if my introduction is good.
Strong prompt (first of several):
Write an 800-word beginner’s guide to baking sourdough at home. Cover starter, schedule, and the three most common beginner mistakes.
Then, in follow-up messages: summarize, translate, brainstorm headlines. Each step builds on a solid foundation instead of competing for attention.
The rule: if your prompt has the word “and” connecting unrelated tasks, split it.
Common mistakes (and quick fixes)
Mistake 1: Asking questions with no goal. “Tell me about marketing” gets you a textbook. “I’m launching a handmade soap shop in Lyon — what are the first 3 marketing steps?” gets you a plan. Always include what you’re trying to achieve.
Mistake 2: Accepting confident nonsense. AI sometimes states wrong facts smoothly. For anything factual — dates, prices, statistics, opening hours — verify against a reliable source. Treat AI as a brilliant brainstorming partner, not a reference book.
Mistake 3: Over-engineering the prompt. Some guides suggest 500-word prompts with elaborate frameworks. For most tasks, that’s overkill. Start simple; add detail only where the answer falls short.
Mistake 4: Forgetting to say what you don’t want. Negative instructions work: “no bullet points,” “don’t use formal language,” “skip the introduction.” Telling the AI what to avoid is often faster than describing what you want.
Mistake 5: Using the same prompt for every tool. ChatGPT, Claude, and Gemini have different strengths and quirks. A prompt tuned for one may underperform on another. When you switch tools, expect a short adjustment period.
Your turn: a 10-minute practice drill
Pick one real task from your life — an email you need to write, a decision you’re weighing, something you’re trying to learn. Write the laziest possible prompt first and read the answer. Then rewrite it using three techniques from this guide: add context, assign a role, set the format. Compare the two answers.
The gap between them is the whole skill. And once you’ve felt it, you’ll never write a lazy prompt again.
Want to put this into practice on a fun project? PromptSkill’s step-by-step guide to planning a trip with ChatGPT is the perfect playground for everything you just learned.