Writing Your First Prompt
You have access to one of the most capable tools ever built -- but it only works as well as your instructions. This lesson teaches you how to write prompts that get useful, accurate, and relevant results on the first try.
The Anatomy of a Good Prompt
Every effective prompt has three core ingredients: context, instruction, and format. Miss any one of these and the AI is left guessing -- which means you get generic, unfocused output. Include all three and the result is dramatically better.
Prompt breakdown
"I run a small online bookshop and I need to send a follow-up email to customers who abandoned their cart."
"Write a friendly, low-pressure email that reminds them what they left behind and offers a 10% discount code."
"Keep it under 150 words. Use a subject line, body, and a clear call-to-action button label."
When you combine all three, the AI knows who you are, what you want, and how you want it delivered. Compare the full prompt above to just saying "write me an email" -- the difference in output quality is enormous.
Think of it this way: context is the background briefing, instruction is the task itself, and format is the delivery specification. You would give all three to a new employee on their first day -- give the same to your AI.
Before and After: Weak vs. Strong Prompts
The gap between a vague prompt and a well-structured one is the gap between useless output and genuinely helpful results. Here are four real-world examples showing exactly how to transform a weak prompt into a strong one.
Writing a professional email
Weak prompt
"Write an email to my boss."
No context about the topic, tone, or purpose. The AI will produce something generic and probably unusable.
Strong prompt
"Write a professional email to my manager requesting to work from home on Fridays. I have been in the role for 2 years with strong performance reviews. Keep the tone respectful but confident. Under 200 words."
Explaining a concept
Weak prompt
"Explain blockchain."
Too broad. You might get a 2,000-word essay or a one-line definition. No way to know.
Strong prompt
"Explain blockchain technology to a 14-year-old who has never heard of it. Use a real-world analogy. Keep it to 3 short paragraphs, no jargon."
Getting code help
Weak prompt
"My code doesn't work."
The AI has no idea what language, what code, or what "doesn't work" means. It will ask you follow-up questions instead of helping.
Strong prompt
"I have a Python function that should return the average of a list of numbers, but it returns 0 for every input. Here is the code: [paste code]. What is wrong and how do I fix it?"
Research and comparison
Weak prompt
"Which laptop should I buy?"
Budget? Use case? Screen size? Operating system preference? The AI has nothing to work with.
Strong prompt
"Compare three laptops under $1,000 that are good for a university student studying graphic design. I need strong display quality and at least 16GB RAM. Present as a table with columns for model, price, display, RAM, and your recommendation."
The Most Common Beginner Mistakes
Nearly everyone makes the same mistakes when they start using AI. Recognising these patterns will immediately improve your results.
1. Being too vague
Prompts like "tell me about marketing" or "help me with my project" give the AI almost nothing to work with. The model has to guess your intent, your audience, your level of expertise, and what format you want. The result is generic output that helps no one.
Fix: Add specifics. What kind of marketing? For what product? What do you already know? What do you need to decide?
2. Not specifying a format
If you do not tell the AI how to structure its response, you will get a wall of text. You might want bullet points but get paragraphs. You might want a table but get a narrative. The AI defaults to whatever seems most natural -- which is usually long-form prose.
Fix: End your prompt with format instructions: "as a numbered list," "in a table," "under 100 words," "as bullet points with bold headings."
3. Not giving context
The AI does not know who you are, what you have already tried, or why you are asking. Without context, it cannot tailor the response. Asking "how do I grow my business?" without mentioning your industry, size, budget, or goals will produce advice so general it could apply to anyone -- and therefore helps no one.
Fix: Start with a sentence or two of background. Who are you? What is the situation? What have you already tried?
4. Asking too many things at once
"Explain quantum computing, compare it to classical computing, list the top companies working on it, and predict where it will be in 10 years." That is four separate tasks crammed into one prompt. The AI will try to do all of them, but each answer will be shallow because it is splitting its effort.
Fix: One prompt, one task. Ask your follow-up questions in the next message. AI conversations are meant to be iterative.
Prompt Patterns That Work
These are proven structures you can reuse across almost any task. Think of them as templates -- fill in your specific details and you will consistently get better results.
Role-setting: "You are a..."
Assigning the AI a specific role or persona dramatically changes the quality and style of its output. When you say "You are a senior nutritionist," the AI adjusts its vocabulary, depth of knowledge, and communication style to match that expertise.
Example
"You are an experienced personal finance adviser. A 28-year-old earning $55,000/year with $12,000 in savings wants to know whether to pay off student loans faster or start investing. Give balanced advice with pros and cons for each option."
Step-by-step instructions
When you need the AI to follow a specific process, break your prompt into numbered steps. This forces the model to think sequentially rather than jumping to a conclusion, which often produces more thorough and accurate results.
Example
"I want to analyse whether I should rent or buy a flat in Manchester. Do this in three steps: 1) List the key financial factors to consider. 2) Compare renting vs. buying using average Manchester prices. 3) Give a recommendation based on someone earning $45,000/year with $20,000 saved."
Examples in the prompt (few-shot learning)
Instead of explaining what you want, show the AI by including one or two examples of the desired output. This is called "few-shot prompting" and it is one of the most effective techniques available. The AI will pattern-match against your examples and produce output in the same style.
Example
"Convert these product features into customer-facing benefits.
Feature: '256GB SSD storage' --> Benefit: 'Loads your apps and files in seconds -- no more waiting around.'
Feature: '18-hour battery life' --> Benefit: 'Work all day on a single charge, no charger needed.'
Now do these:
Feature: '6.1-inch OLED display'
Feature: 'IP68 water resistance'
Feature: '48MP triple camera system'"
Specifying output format
Telling the AI exactly how to structure its response saves you from reformatting later. You can request tables, bullet points, JSON, markdown, numbered lists, or any structure you need. Be explicit -- the AI will follow your format instructions precisely.
Example
"List 5 healthy weeknight dinner ideas. For each one, provide: the meal name in bold, the estimated cooking time in parentheses, a one-sentence description, and a difficulty rating of Easy, Medium, or Hard. Format each entry as a numbered item."
For a deeper dive into these techniques -- including chain-of-thought reasoning, system prompts, and temperature settings -- see the Advanced Prompting Techniques guide.
Your First 10 Prompts
Here are 10 ready-to-use prompts covering different tasks. Copy any of them, paste into your preferred AI tool, and see the results. Each one demonstrates the context + instruction + format structure in action.
"Summarise the following article in exactly 3 bullet points. Each bullet should be one sentence and capture a key takeaway. Write for someone who has 30 seconds to read this. [Paste article text here]"
"Explain how compound interest works to someone with no financial background. Use a concrete example with real numbers showing what happens to $1,000 over 10 years at 5% annual interest. Keep it under 200 words."
"Compare React and Vue.js for a beginner who wants to build their first web application. Create a table with columns for: Learning Curve, Community Size, Job Market, and Best For. Add a one-sentence recommendation at the end."
"Write a LinkedIn post announcing that I just completed a data analytics certification. I want to thank my employer for supporting my development. Tone should be genuine and professional, not overly enthusiastic. Under 150 words, no hashtags."
"Analyse the strengths and weaknesses of this business idea: a subscription box for home-brewed coffee that delivers freshly roasted beans, a recipe card, and a small snack each month for $25. List 3 strengths and 3 risks as bullet points."
"Brainstorm 10 blog post titles for a personal finance website aimed at people in their 20s. The tone should be approachable and slightly irreverent -- not stuffy or corporate. Each title should be under 10 words."
"Translate the following English text into Spanish, then provide the translation a second time with each sentence annotated to explain any tricky grammar or vocabulary choices for an intermediate Spanish learner. Text: 'We would have arrived on time if the train hadn't been cancelled at the last minute.'"
"Here is a JavaScript function that should filter an array of products to only show items under $50, but it returns an empty array every time. Find the bug, explain why it happens, and give me the corrected code. [Paste your code here]"
"Create a 4-week study plan for someone preparing for the AWS Cloud Practitioner certification. They can study 1 hour per day on weekdays. Break it into weekly goals with specific topics and recommended resources. Format as a table with columns for Week, Topics, Daily Tasks, and Resources."
"Review the following cover letter for a junior marketing role. Be honest and specific. Point out any cliches, vague claims, or weak sections. Then suggest how to fix each issue. Do not rewrite the whole letter -- just list the problems and fixes. [Paste cover letter here]"
For hundreds more task-specific prompts organised by model and use case, visit the Prompt Library.