How to Write a Prompt That Gets You What You Want — From Vague Request to Precise Instruction
“Write me an article about tourism” — that’s not a request, it’s a fog signal.
The difference between people who get exceptional results from AI and those
who close the app in frustration isn’t intelligence — it’s the prompt.
A practical guide with real, copy-ready examples, taking you from vague
requests to precise instructions that actually work.
A prompt is the difference between an assistant that understands you and one that wastes your time. Writing one well is a skill you can learn in an hour — and master with practice.
Why Most People Fail at the Start
When someone sits in front of an AI program for the first time, they type something like: “Write me an article about tourism.” Then they read the output, frown, and close the app convinced that “AI doesn’t understand my field” or “these tools aren’t useful for real work.”
The problem isn’t the program. The problem is the request.
“Write me an article about tourism” is exactly like walking into a restaurant and telling the waiter: “Bring me food.” The chef is there, the ingredients are there — but no one knows what you want.
A prompt — the instruction you give an AI model — is the bridge between what you want and what the AI produces. In this article we teach you how to build one correctly, step by step, with real examples you can copy directly.
The Core Principle: AI Excels at Execution and Needs Context
An AI model is not a mind that thinks with spontaneous initiative — it’s an exceptional execution engine. It gives you what you asked for precisely, not what you need intelligently. So the more context you provide, the higher the quality of the output.
Think of it this way: if you hired a new assistant on their first day and said “handle my email,” what would they do? They’d act generically and predictably. But if you said: “I’m a translator specializing in legal texts, most of our clients are Gulf companies, my style is formal and direct, responses should go out within two business days” — they’d perform like a professional.
That’s exactly what a good prompt does.
The Basic Structure of an Effective Prompt: Four Elements
Not every prompt needs all four elements — but understanding them lets you know which to add depending on the situation.
1. Role
Tell the program who it is in this conversation. “You are a professional translator specializing in legal texts” is completely different from “You are an Arabic language instructor” or “You are a seasoned news editor.” Defining the role calibrates the tone of the answer, its technical level, and the type of vocabulary used.
2. Task
What exactly do you want done? Be as specific as possible. “Write” is weaker than “write an educational article.” And “write an educational article” is weaker than “write an educational article for beginners explaining how to create an account on Upwork.”
3. Context
Information that helps the program understand the situation: Who is your audience? What is the purpose of the text? What are the constraints? What style is required?
4. Format
How do you want the output? Numbered list or paragraphs? Short or detailed? With subheadings or without? What language? Approximately how long?
From Theory to Practice: Real Examples
This is the most important section of this article. We’ll take weak prompts and transform them into strong ones, explaining what changed each time.
Example 1 — Content Writing
Weak:
“Write an article about the benefits of coffee.”
Strong:
“You are a content writer specializing in health and nutrition. Write an educational article in formal Arabic about the benefits and risks of coffee, aimed at readers in their thirties who care about their health. Style: direct and practical, not academic. Length: 600–800 words. Start with a surprising statistic or fact that grabs attention. Divide it into sections with subheadings.”
Example 2 — Translation
Weak:
“Translate this text into Arabic.” [+ text]
Strong:
“You are a professional English-Arabic translator. Translate the following text into formal Arabic with these considerations: preserve the formal tone of the original. Do not translate company or product names — keep them in English. When a technical term appears, write it in Arabic followed by the English term in parentheses. Return the translation only, without any commentary.” [+ text]
Example 3 — Editing
Weak:
“Improve this text.” [+ text]
Strong:
“You are a professional literary editor. Review the following text and: correct grammatical and spelling errors. Improve sentence flow while preserving the author’s original style. Suggest alternatives for any sentence that feels repetitive or flat. Do not add new content. Return the full edited text, then a list of the changes you made with a brief explanation of each.”
Five Techniques That Improve Your Prompt Immediately
First: Say What You Don’t Want
Models tend toward length, filler, and long introductions. If you don’t want that, say so explicitly: “Don’t start with a general introduction about the importance of this topic. Start directly with the first piece of information.” Or: “Don’t use phrases like: of course, certainly, great question, happy to help.”
Second: Give It an Example of What You Want
Instead of describing the required style, attach an example: “Write in a style similar to this paragraph: [paragraph].” Models are excellent at replicating patterns and styles.
Third: Ask for Multiple Options
Instead of requesting one result that might not satisfy you: “Give me three different options for this article’s title — the first serious, the second a question, the third surprising or unexpected.”
Fourth: Work in Stages
For large tasks, don’t send everything at once. Break the work down: ask for the structure first, review it, then ask for each section in detail. This gives you greater control over the final output.
Fifth: Ask It to Critique Its Own Work
After it produces a text, add: “Now review what you wrote and tell me: what is its weakest point? And what could have been said better?” This sometimes reveals problems in the output it would have hidden if you hadn’t asked.
For Professionals: Advanced Techniques Worth Investing In
System Prompt — The Fixed Instruction
Most programs allow writing “system instructions” that apply to every new conversation. Instead of repeating your context each time, write it once: “I am a professional English-Arabic translator specializing in legal texts, working with Gulf market clients, formal style…” And every conversation starts with the program already knowing you.
In Claude: Settings → User Preferences. In ChatGPT: Settings → Customize ChatGPT.
Chain of Thought — Step-by-Step Reasoning
For complex analytical tasks, add at the end of your prompt: “Think step by step before you answer.” This simple sentence meaningfully improves analysis quality by forcing the model to build its conclusions methodically.
Few-Shot Prompting — Learning From Examples
Give it two or three examples of what you want before the actual task:
“Here’s how I want headlines phrased:
Example A: ‘Machine Translation: What It Gets Right and Where It Falls Short’
Example B: ‘Claude or ChatGPT — Which One for Your Work?’
Now write five headlines in the same style for an article about time management in freelancing.”
We detail these techniques in a practical applied context in our article Using Claude for Arabic Content Writing.
Common Mistakes — and How We Avoid Them
Mistake 1: The vague open-ended request. “Write something creative about Ramadan” gives the model no direction. Always specify: what form? what audience? what purpose?
Mistake 2: Contradictory requirements. “Write something short and detailed and comprehensive” — brevity and comprehensiveness don’t coexist. Choose your priorities.
Mistake 3: Accepting the first output without review. Prompting is a dialogue, not a one-time command. If the output doesn’t satisfy you, refine and try again: “The output is too long — cut it in half while keeping the core ideas.”
Mistake 4: Blind trust in statistics and figures. AI models generate numbers that sound precise but are sometimes invented. Any figure or statistic in the output needs verification from an independent source. We address this specifically in our article What AI Cannot Do — Limits You Must Know.
Ready-to-Copy Prompts — A Practical Library
These are prompts we use regularly — copy and adapt them to your needs:
For Translators:
“You are a professional [Language A] to [Language B] translator. Translate the following text with these considerations: [required style]. Technical terms: [translate them / keep them in original]. Target reader: [audience description]. Return only the translation, no commentary.”
For Content Writers:
“Write an article in [language] about [topic]. Audience: [description]. Length: [number] words. Style: [formal / conversational / technical / narrative]. Start with [a question / a statistic / a short story]. Add subheadings. Don’t add a general introduction about the topic’s importance.”
For Editors:
“Review the following text and correct: spelling and grammatical errors. Sentences that are too long: suggest splitting them. Unintentional repetition: identify it. Preserve the author’s style. Return the corrected text followed by a summary of changes.”
For Freelancers — Email Replies:
“Write a professional reply to the following email. Situation: [brief description of what you want to say]. Tone: friendly but professional. Length: no more than two paragraphs. Don’t start with ‘I hope this email finds you well.'”
For more practical applications in a freelancing context, see our article AI for Freelancers — 10 Tasks in Half the Time.
The Takeaway: A Good Prompt Is the First Half of the Work
Everything in this article can be summarized in one rule: the more context you give the program, the more precise the output it gives you back.
Don’t expect a perfect result on the first attempt — prompting is a dialogue, and every response teaches you how to improve your next request. Professionals who get the best results from these tools don’t have a secret — they simply write clearer, more specific prompts.
Next step: now that you know how to ask, it’s equally important to know where these programs hit their limits — what AI doesn’t do well, no matter how precisely you phrase your prompt. That’s what the next article in this series covers: What AI Cannot Do — Limits You Must Know.
