Most people get mediocre results from AI not because the model is weak, but because the prompt is vague. AI is a prediction engine. The more clearly you specify context, goal, and constraints, the more accurately it can predict the answer you actually want.

Think of prompting as briefing a smart intern: capable, fast, but needs context and has no idea what you want unless you say it.

The Core Formula

A strong prompt has 5 parts, not always all needed, but useful to check:

1. Role + Context Who should the AI act as, and what background should it know?

2. Task What exactly should it do?

3. Constraints What are the limits: length, tone, format, audience, what to avoid?

4. Examples (if needed) Show what good looks like.

5. Success Criteria How will you know it’s good?

Weak: “Write about crochet hooks” Strong: “You are a crochet instructor for beginners. Write a 180-word guide explaining when to use a 5.0 mm vs 6.0 mm hook with worsted yarn. Audience is absolute beginners. Tone: friendly and practical. Include: 1) how hook size affects fabric drape, 2) one simple gauge tip. Format: short paragraphs with bullet points. Avoid jargon.”

7 Practical Techniques That Actually Work

1. Be Specific About Output Format

AI defaults to long generic essays. Tell it the shape you want.

Try:

• “Give me a table with columns: Task, Time Estimate, Tools Needed”
• “Answer in 5 bullets, each under 15 words”
• “First give a 1-sentence answer, then 3 supporting details”

This saves editing time and makes results immediately usable.

2. Give Context and Audience

The same information for a beginner vs an expert should sound completely different.

Instead of: “Explain single crochet vs double crochet”

Use: “Explain single crochet vs double crochet for someone who learned to chain yesterday and wants to make a scarf. Use analogies, avoid abbreviations.”

Add relevant constraints: “I have only 30 minutes a day, I work on my phone, I am in Guelmim with limited yarn options.”

3. Show, Don’t Just Tell (Few-Shot Prompting)

If you want a certain style, give 1-2 examples.

Example: “Rewrite captions in this style: Example 1: ‘Cozy cotton, made slow on a rainy afternoon.’ Example 2: ‘Beige yarn, wooden hook, quiet hands.’ Now rewrite this caption: ‘Handmade bag made of cotton'”

This is far more effective than saying “be aesthetic.”

4. Break Complex Tasks Into Steps

Don’t ask for a full business plan in one prompt. Chain prompts.

Step 1: “List 10 low-cost product ideas for a crochet shop in Morocco using locally available cotton.” Step 2: “Pick the top 3 based on speed to make and shipping ease, explain why.” Step 3: “For idea #1, draft a product description, 120 words, SEO-friendly.”

Models perform better on one focused task than on five tasks at once.

5. Assign a Role

Role prompting focuses the model’s knowledge.

• “Act as a patient crochet teacher who explains mistakes gently”
• “Act as a strict editor who cuts fluff and fixes grammar only”
• “Act as a Moroccan artisan market seller who knows what tourists look for”

It doesn’t give the model new abilities, but it narrows which patterns it pulls from.

6. Ask It to Think Before It Answers

For reasoning, planning, or comparison tasks, prompt it to show its work.

Add: “Think step by step. List the pros and cons first, then give your final recommendation.”

Or: “Before answering, ask me 2 clarifying questions if my request is ambiguous.”

This reduces hallucinations and surfaces assumptions.

7. Iterate and Constrain Negatives

Your first prompt is rarely perfect. Treat it as a conversation.

Follow-up prompts that help:

• “Make it shorter, remove adjectives”
• “Keep the structure but make it for WhatsApp, more casual”
• “This is too generic, add one concrete example from everyday life”
• “Don’t use em dashes, don’t start with ‘In today’s world'”

Use negative constraints explicitly: “Do not include prices, do not mention acrylic yarn.”

Common Prompt Mistakes

Too vague: “Help me with AI tools” -> AI doesn’t know if you want business use, learning, or creative use.

Too much at once: “Write a blog post, create images, give me SEO keywords, and make a marketing plan” -> Do one at a time.

No audience: Writing for yourself, a customer, or a child needs completely different language.

Forgetting to verify: Never paste unverified facts, citations, or statistics directly into a final document. Add: “If you are unsure about a fact, say ‘I don’t know’ and suggest where to check.”

Prompt Templates You Can Reuse

For learning: “Teach me as if I am a beginner who knows [what you already know]. Explain in 3 short parts, give one analogy, then quiz me with 2 questions.”[topic]

For writing: “Context: [who you are, who it’s for]. Task: Write about. Tone: [friendly/professional/etc]. Length:. Must include:. Must avoid:. Format: [bullets/table/paragraphs].”[type][topic][words][points][things]

For summarizing: “Summarize this text for. Give: 1) One-sentence main idea, 2) 3 key facts, 3) What action should I take? Text: “[audience][paste]

For decisions: “I need to choose between [Option A] and [Option B]. My constraints are. List pros/cons of each for my situation, then recommend one with reasoning.”[time][budget][skills]

Final Tip: Save Your Best Prompts

When you get a great result, save the prompt itself. Build a small personal library in your notes app: prompts for email replies, product descriptions, lesson plans, etc. Over time, your prompts become reusable tools, not one-off questions.

Good prompting is not about tricking the AI. It is about being clear about what you want, who it’s for, and what good looks like. Do that, and the results become much more useful with less back-and-forth.