The Art of Asking: Prompting Basics

• Understand that better prompts get better results
• Know the four elements of a good prompt: context, task, format, constraints
• Write prompts that are specific enough to be useful
• Recognize vague prompts and how to improve them
• Stop treating AI like a search engine

You can use AI now. But are you getting good results?

The difference between disappointing AI output and genuinely useful AI output almost always comes down to one thing: how you ask. This is called "prompting," and it is the most important skill you will learn in this course.

The good news: prompting is not complicated. It follows predictable patterns. Once you understand them, you will immediately get better results.

Here is the fundamental truth of working with AI:

The quality of your output is directly proportional to the quality of your input.

AI does not read minds. It cannot guess what you really meant. It responds to exactly what you wrote, nothing more.

A Simple Example

Watch what happens with the same basic request phrased two different ways:

Vague prompt: "marketing help"

Specific prompt: "Write three email subject lines for a 20% off sale on running shoes. Target audience: people who signed up after seeing our marathon training guide. Tone: energetic but not pushy."

The first prompt might get you a generic explanation of marketing concepts. The second gets you three usable subject lines you can actually send.

Same AI. Same task. Dramatically different results.

Most effective prompts include four components. You do not need all four every time, but knowing them helps you diagnose why a prompt is not working.

1. Context (Who and Why)

Tell AI the relevant background:

• Who are you in this situation?
• Who is the audience?
• What led to this request?

Without context: "Write an apology"

With context: "I am a customer service manager. A shipment arrived 2 weeks late to an important client. Write an apology email that acknowledges the delay without making excuses."

2. Task (What Exactly)

Be specific about what you want AI to do:

• Create? Explain? Compare? Summarize? Review? Brainstorm?
• The verb matters more than you think

Unclear task: "Tell me about solar panels"

Clear task: "Compare the three most popular residential solar panel brands. Include pros, cons, and typical costs for a 2,000 square foot home."

3. Format (How It Should Look)

Specify the structure you want:

• Bullet points or paragraphs?
• How long?
• What sections?
• What style?

No format guidance: "Explain machine learning"

With format guidance: "Explain machine learning in exactly 3 paragraphs. First paragraph: what it is. Second paragraph: how it works at a high level. Third paragraph: one real-world example. Use simple language a high schooler would understand."

4. Constraints (What to Avoid)

Tell AI what NOT to do:

• Topics to skip
• Tone to avoid
• Length limits
• Assumptions not to make

Without constraints: "Write a product description for our new headphones"

With constraints: "Write a product description for our new headphones. 100 words maximum. Do not mention competitor brands. Avoid technical jargon. Do not claim they are the best or use superlatives."

The biggest mindset shift for new AI users: AI is a conversation, not a search query.

Search Engine Thinking (Avoid This)

• Short keyword phrases
• One question, one answer
• Start over if results are bad
• Cannot ask follow-up questions

Conversation Thinking (Do This)

• Full sentences with context
• Back-and-forth refinement
• Build on previous responses
• Ask for clarification, expansion, revision

Search query style (weak): "best laptop 2026"

Conversation style (strong): "I need a laptop for video editing and occasional gaming. Budget is $1,500. I travel a lot so battery life matters. I currently use a MacBook but am open to Windows. What should I consider?"

The second approach treats AI as a knowledgeable assistant, not a search index.

Not every prompt needs all four elements. Short prompts work when:

• The task is common and obvious: "What is the capital of France?"
• You are brainstorming and want variety: "Give me 10 business name ideas"
• You plan to iterate: "Draft a meeting agenda" (then refine in follow-ups)
• Context is genuinely unnecessary: "Translate this to Spanish: Hello, how are you?"

The goal is not to write long prompts. The goal is to give AI enough information to give you useful output on the first try.

• Input quality equals output quality: Better prompts get dramatically better results
• Use the four elements: Context, Task, Format, Constraints
• Be specific but not verbose: Say enough to be useful, not more
• Think conversation, not search: AI can handle full sentences and follow-ups
• Simple prompts work for simple tasks: Do not overcomplicate when unnecessary

Take a vague prompt and transform it using the four elements.

Start with this vague prompt: "help me write an email"

Now expand it:

1. Add context: Who is this to? What is the situation?
2. Add task: What kind of email? What should it accomplish?
3. Add format: How long? What tone?
4. Add constraints: Anything to avoid?

Write your expanded prompt. Then try both versions (vague and specific) with AI.

Compare the results. The difference will be obvious.

Example transformation: "Write a follow-up email to a job interviewer. I interviewed for a marketing coordinator position at a nonprofit last Tuesday. The interviewer was Sarah Chen. Keep it under 100 words. Professional but warm. Thank her for her time without being overly formal or desperate."

Now you understand why prompting matters.

• Anthropic prompt engineering guide: https://docs.anthropic.com/en/docs/build-with-claude/prompt-engineering
• OpenAI prompt engineering best practices: https://platform.openai.com/docs/guides/prompt-engineering
• Research on prompt design: https://arxiv.org/abs/2302.11382