Look, I’m just going to say it: most people are terrible at writing AI prompts.
Not because they’re dumb or lazy—far from it. It’s because nobody really teaches this stuff. We all just sort of stumbled into the AI revolution, got handed these powerful tools, and were basically told “good luck, figure it out.” It’s like being given the keys to a Ferrari when you’ve only ever driven a golf cart.
But here’s the thing: the difference between a mediocre AI prompt and a great one isn’t rocket science. You don’t need a computer science degree or some secret prompting handbook that only the tech elite have access to. You just need to understand a few key principles—principles that I’m about to share with you.
After spending way too many hours experimenting with ChatGPT, Claude, and other AI tools (and making basically every mistake possible), I’ve figured out what actually works. So let’s dive into the 9 rules that will transform you from an AI prompt amateur into someone who gets consistently amazing results.
Rule #1: Be Ridiculously Specific (No, More Specific Than That)
Remember when you were a kid and you’d ask your parents for “something to eat” and they’d give you a carrot? Yeah, AI is kind of like that.
The biggest mistake people make with AI prompts is being vague. They’ll write something like “write me a blog post about marketing” and then wonder why the output is generic, boring, and sounds like it was written by a robot (which, fair enough, it was).
Here’s the difference:
Bad prompt: “Write about social media marketing.”
Good prompt: “Write a 1,500-word blog post about Instagram marketing strategies for small coffee shops, focusing on three specific tactics: behind-the-scenes content, user-generated content, and limited-time offer promotions. Use a friendly, conversational tone and include specific examples.”
See the difference? The second prompt gives the AI a clear target to hit. It knows the length, the audience, the specific tactics to cover, the tone to use, and even asks for examples. The AI doesn’t have to guess what you want—you’ve told it exactly what you’re looking for.
Think of it this way: AI is like a really talented but slightly clueless intern. They’re capable of amazing work, but they need clear instructions. The more specific you are, the better their output will be.
Let me give you another example to hammer this home. Let’s say you’re trying to write an email:
Vague prompt: “Write an email to my team about the meeting.”
The AI has no idea what meeting, what the purpose is, what tone to use, or what you want people to do. It’ll probably give you something generic that you’ll have to completely rewrite.
Specific prompt: “Write a 150-word email to my 8-person product team reminding them about tomorrow’s quarterly planning meeting at 2 PM in Conference Room B. Mention that they should review the Q4 goals document I sent last week and come prepared with their top 3 priorities for Q1. Keep the tone professional but friendly, and end with a note that I’m bringing coffee and donuts.”
Now the AI has everything it needs: audience, purpose, length, key details, action items, tone, and even a nice personal touch at the end.
Here’s what specificity looks like across different use cases:
For writing: Instead of “write a story,” try “write a 1,000-word suspense story about a detective who realizes the missing person they’re searching for is watching them. Use first-person perspective, present tense, and end with a cliffhanger.”
For analysis: Instead of “analyze this data,” try “analyze this sales data and identify the top 3 trends over the past 6 months. Focus on seasonal patterns, product category performance, and geographic variations. Present findings as bullet points with specific percentages.”
For learning: Instead of “explain quantum physics,” try “explain quantum superposition to someone who understands basic chemistry but hasn’t studied physics since high school. Use an analogy involving everyday objects and avoid equations.”
Pro tip: If you catch yourself using words like “some,” “a few,” or “things,” you’re probably not being specific enough. Challenge yourself to replace those vague terms with concrete details. Instead of “write some social media posts,” say “write 5 Instagram captions.” Instead of “give me a few ideas,” say “give me 10 specific ideas, each with a brief explanation.”
The magic formula for specificity is: What + Who + Why + How + Any Constraints. Answer those five questions in your prompt, and you’re already miles ahead of most people.
Rule #2: Give the AI a Role to Play
This one’s a game-changer that most people overlook.
When you give the AI a specific role or persona, it completely transforms the quality and style of the output. It’s like method acting for artificial intelligence.
Instead of just asking for information, try this:
“You are a senior software engineer with 15 years of experience. Explain how APIs work to a complete beginner who’s never coded before.”
Or:
“You are a personal trainer specializing in strength training for busy professionals. Create a 30-minute workout routine that can be done at home with minimal equipment.”
Why does this work? Because by defining a role, you’re activating specific knowledge patterns and communication styles. A senior engineer explains things differently than a college professor. A personal trainer writes differently than a doctor. And the AI can tap into these different perspectives.
I’ve used this trick for everything from getting marketing advice (“You are a CMO of a successful B2B SaaS company…”) to learning complex topics (“You are an enthusiastic high school teacher who makes physics fun…”). The results are consistently better than generic prompts.
Here’s why the role-playing technique is so powerful: it primes the AI to access specific knowledge frameworks and communication styles. It’s like telling an actor what character they’re playing—suddenly, they know how to move, speak, and think.
Some of my favorite roles to assign:
For business advice: “You are a business consultant who has helped 50+ startups achieve product-market fit. You’re practical, data-driven, and allergic to buzzwords.”
For technical explanations: “You are a senior developer who’s known for explaining complex technical concepts in ways that non-technical stakeholders can understand.”
For creative work: “You are an award-winning copywriter who specializes in writing headlines that stop people mid-scroll.”
For problem-solving: “You are a troubleshooting expert who approaches every problem methodically, always considering the obvious solutions first before exploring complex ones.”
For editing: “You are a ruthless editor who cuts unnecessary words and strengthens every sentence. Your motto is ‘brutal clarity.’”
You can even combine roles for more nuanced outputs: “You are both a nutritionist and a busy parent of three. Create a weekly meal plan that’s healthy but realistic for someone with limited time and picky eaters.”
The role doesn’t have to be a job title, either. You can say “You are someone who’s passionate about sustainability and approaches every topic through that lens” or “You are skeptical by nature and always look for potential downsides or risks.”
One warning though: don’t just use roles randomly. Match the role to what you actually need. If you want creative, punchy copy, don’t ask the AI to roleplay as a lawyer. If you need careful, precise technical documentation, don’t ask it to be a comedian.
Rule #3: Provide Examples (Show, Don’t Just Tell)
You know how when you’re trying to explain something, you often say “like this” and show an example? AI responds really well to the same approach.
Let’s say you want the AI to write product descriptions in a particular style. Don’t just describe the style—show it:
“Write product descriptions in this style:
Example 1: ‘This isn’t just a mug. It’s your morning motivation, your 3 PM pick-me-up, and your excuse to take a five-minute break. Made from ceramic that actually keeps your coffee hot, it holds 12 oz of whatever gets you through the day.’
Example 2: ‘Meet your new favorite notebook. 200 pages of possibility, wrapped in a cover that won’t fall apart when you throw it in your bag. Because your brilliant ideas deserve better than some flimsy spiral thing from the drugstore.’
Now write a description for a desk lamp.”
By showing examples, you’re essentially training the AI on the fly. It picks up on tone, structure, length, and style in ways that are hard to explain with words alone.
This is especially powerful for:
- Writing in a specific brand voice
- Matching a particular format or structure
- Replicating a style of communication
- Creating content that fits with existing materials
Rule #4: Break Complex Tasks Into Smaller Steps
Imagine asking someone to “plan a wedding” versus giving them a checklist of specific tasks. Which approach do you think would work better?
Same deal with AI.
When you have a complex project, don’t dump it all in one massive prompt. Break it down into logical steps and tackle them one at a time.
Instead of: “Create a complete content marketing strategy for my business.”
Try this sequence:
-
First prompt: “Help me identify my target audience. My business is [describe business]. Ask me questions to understand who we should be targeting.”
-
Second prompt: “Based on [audience details], what are the main pain points and challenges this audience faces?”
-
Third prompt: “Given these pain points, what types of content would be most valuable to create? Suggest 10 specific content ideas.”
-
Fourth prompt: “Let’s develop a content calendar. Take these ideas and organize them into a 3-month publishing schedule…”
See how much more manageable that is? Plus, you get to review and adjust at each stage instead of getting one giant output that might not be quite right.
This approach is especially useful for:
- Research projects
- Planning and strategy work
- Creative writing with multiple elements
- Technical problem-solving
- Any task where you might need to pivot based on initial results
Think of it like cooking a complex meal—you don’t try to do everything simultaneously. You prep, then cook components in order, then bring it all together.
Here’s another example of this in action. Let’s say you’re writing a business proposal:
Single massive prompt (doesn’t work well): “Write a complete business proposal for implementing a new customer service chatbot system.”
Multi-step approach (works much better):
Step 1: “Help me outline the key sections that should be in a business proposal for implementing a chatbot system. What are the critical components?”
Step 2: “Great, now let’s work on the executive summary. Based on [details you provide], write a one-page executive summary hitting the key points.”
Step 3: “Now let’s detail the problem statement. Here’s what I know about our current customer service challenges: [details]. Expand this into a compelling problem statement.”
Step 4: “Based on the problem statement, create a proposed solution section that addresses each challenge specifically.”
And so on. Each step builds on the previous one, and you maintain control over the direction.
The beauty of this approach is that you can course-correct at any stage. If Step 2 goes in the wrong direction, you fix it before investing time in Steps 3, 4, and 5. It’s iterative and adaptive.
This also works brilliantly for learning complex topics:
Step 1: “Explain the basics of [topic] in simple terms.” Step 2: “Now explain [specific subtopic] in more detail.” Step 3: “What are common misconceptions about [topic]?” Step 4: “Give me three real-world examples of [topic] in action.” Step 5: “What should I learn next to deepen my understanding?”
Each prompt builds your knowledge progressively, like climbing a ladder instead of trying to leap to the top in one jump.
Pro tip: If you’re tackling a really big project, keep a separate document where you track your prompts and the AI’s responses. This becomes your “prompt log” and helps you maintain continuity across a long session or multiple sessions.
Rule #5: Use Constraints to Your Advantage
Here’s a counterintuitive truth: limitations often lead to better results.
When you give AI complete freedom, it tends to produce generic, middle-of-the-road content. But when you add specific constraints, it gets creative within those boundaries—and the output becomes more interesting and useful.
Try adding constraints like:
Length: “Write this in exactly 280 characters” or “Keep it under 100 words”
Format: “Structure this as a numbered list with exactly 7 items” or “Write this as a dialogue between two people”
Restrictions: “Explain this without using jargon” or “Describe this product without mentioning price”
Audience level: “Write this for someone with zero background knowledge” or “Assume the reader is an expert”
Tone requirements: “Keep it professional but not stuffy” or “Make it funny without being cheesy”
Here’s a real example:
“Write an email to a client explaining that we’ll miss the deadline. Keep it under 150 words, maintain a professional apologetic tone, offer a specific solution, and end on a positive note about the relationship.”
Those constraints force the AI to be thoughtful and precise. Without them, you might get a rambling email that hits some points but misses others.
Constraints aren’t limitations—they’re guardrails that help the AI stay focused on what actually matters for your specific need.
Rule #6: Tell the AI What NOT to Do
Sometimes the best way to get what you want is to explicitly state what you don’t want.
This is especially useful when AI tends to fall into certain patterns that don’t work for your situation.
For example:
“Write a welcome email for new subscribers. DO NOT use phrases like ‘we’re thrilled,’ ‘excited to have you,’ or ‘can’t wait.’ DO NOT include emojis. DO NOT make it longer than 100 words.”
Or:
“Create a job description for a graphic designer. DO NOT use buzzwords like ‘rockstar,’ ‘ninja,’ or ‘guru.’ DO NOT list unrealistic requirements. DO NOT make the tone overly casual or overly formal.”
This negative prompting technique helps you avoid common AI clichés and patterns. You know how AI sometimes sounds a bit too… AI-ish? This helps cut through that.
Some common things you might want to tell AI NOT to do:
- Don’t be overly enthusiastic or use excessive exclamation points
- Don’t use clichés or overused phrases
- Don’t make assumptions about [specific thing]
- Don’t include filler words or fluff
- Don’t write in a corporate/robotic tone
- Don’t give general advice—be specific
It’s like working with a really enthusiastic helper who tends to go overboard. A gentle “hey, not that” helps them calibrate their output to your actual preferences.
Rule #7: Ask the AI to Think Step-by-Step
This might sound weird, but one of the most powerful prompting techniques is literally asking the AI to show its work.
When you include phrases like “think through this step-by-step” or “explain your reasoning,” you often get dramatically better results. Why? Because it forces the AI to process the information more thoroughly rather than jumping to conclusions.
Compare these two approaches:
Basic prompt: “Should I invest in paid advertising or content marketing for my startup?”
Step-by-step prompt: “I’m deciding between paid advertising and content marketing for my startup. Think through this step-by-step:
- What factors should I consider?
- What are the pros and cons of each approach?
- What questions do you need answered to give a better recommendation?
- Based on typical startup scenarios, what would you suggest and why?”
The second prompt leads to much more thoughtful, nuanced responses. The AI doesn’t just give you a quick answer—it actually works through the problem methodically.
This technique is incredibly valuable for:
- Problem-solving and decision-making
- Complex analysis
- Learning and education
- Debugging code or troubleshooting issues
- Strategic planning
You can also use variations like:
- “Walk me through your thought process”
- “Explain how you arrived at this answer”
- “Break down the logic behind your recommendation”
- “Show your work”
It’s the difference between asking someone for an answer and asking them to explain how they got that answer. The latter is almost always more useful.
Rule #8: Iterate and Refine (It’s Not a One-Shot Game)
Here’s something nobody tells beginners: your first prompt usually won’t be your best.
Great AI interactions are more like conversations than single commands. You get an initial output, see what works and what doesn’t, then refine from there.
Don’t be afraid to follow up with:
- “Make it more [specific quality]”
- “That’s good, but can you make it shorter/longer?”
- “Keep the structure but change the tone to be more [X]”
- “I like paragraphs 2 and 3, but rewrite paragraph 1 to focus on [Y]”
- “This is close, but let’s adjust…”
Think of it like working with a collaborator. You wouldn’t expect a colleague to nail exactly what you want on the first try, right? Same principle here.
I’ve had sessions where my tenth prompt produced something amazing, but prompts 1-9 were necessary steps to get there. Each iteration helped me (and the AI) understand what I actually wanted.
Here’s what effective iteration looks like in practice:
First attempt: “Write a LinkedIn post about productivity tips.”
AI gives you something generic. So you iterate:
“That’s a good start, but let’s make it more specific. Focus on productivity tips specifically for remote workers who struggle with work-life boundaries. Make it more personal and less listicle-y.”
Getting closer, but not quite there:
“Better! Now can you remove the first paragraph entirely and start with the second paragraph? Also, that last tip about ‘setting boundaries’ is too vague—replace it with a specific, actionable technique.”
Almost perfect:
“Perfect structure now. Last thing—can you rewrite the opening hook to be more attention-grabbing? Maybe start with a relatable pain point instead of a statement.”
See how each iteration gets more precise? You’re not starting from scratch each time; you’re refining and adjusting. It’s like sculpting—you start with a rough shape and gradually add detail until it’s exactly what you want.
Some of my favorite iteration prompts:
- “That’s good but make it 50% shorter without losing the key points”
- “Keep the facts but make the tone more [conversational/formal/funny/serious]”
- “This is almost perfect—just rewrite the conclusion to be more [specific quality]”
- “Can you give me three different versions of that headline?”
- “Expand point #3 with a specific example”
- “Remove any jargon and explain this as if I’m a beginner”
The iteration process also teaches you what works. Over time, you’ll notice patterns: certain types of prompts consistently produce better results, specific phrasings get you closer to what you want, and some approaches work better for certain tasks.
Pro tip: Keep track of what works. When you stumble on a prompt formula that produces great results, save it. Build your own library of effective prompts that you can reuse and adapt. I have a notes document with my “greatest hits” prompts that I reference constantly. It saves me tons of time and gives me a starting point that I know works.
The best prompt writers aren’t the ones who write perfect prompts every time—they’re the ones who know how to iterate effectively. They understand that prompting is a conversation, not a command.
Rule #9: Context is Your Secret Weapon
Last but definitely not least: give the AI context.
AI doesn’t know anything about you, your business, your audience, or your specific situation unless you tell it. And the more context you provide, the more tailored and useful the output becomes.
Bad prompt: “Write a social media post about our new product.”
Better prompt: “Write a social media post about our new project management tool. We’re targeting small marketing agencies (5-20 person teams) who are frustrated with Asana’s complexity and Monday.com’s cost. Our main selling point is simplicity plus power—easy to learn but with advanced features when you need them. Our brand voice is friendly and straightforward, never condescending. Keep it under 100 words.”
See all that context? That’s gold. It helps the AI understand:
- What you’re selling
- Who you’re selling to
- What makes you different
- How you communicate
- What constraints matter
The more context you provide up front, the less back-and-forth you’ll need later.
Good context to include:
- Your audience and their characteristics
- The purpose or goal of the content
- Your brand voice and values
- Relevant background information
- Any constraints or requirements
- What success looks like
Don’t make the AI guess. It’s like the difference between telling a taxi driver “downtown” versus giving them a specific address—one gets you sort of where you want to go, the other gets you exactly there.
Putting It All Together
Alright, let’s see these rules in action with a real example.
Weak prompt: “Write a guide about email marketing.”
Strong prompt using our rules: “You are an email marketing consultant who specializes in e-commerce (Rule #2: Role). Write a practical, actionable guide about abandoned cart emails for online stores selling products in the $50-200 range (Rule #1: Specificity).
Structure it as 5 specific strategies, each 100-150 words (Rule #5: Constraints), with a concrete example for each strategy (Rule #3: Examples). Think through what actually makes customers return to complete their purchase, considering both psychological triggers and practical concerns (Rule #7: Step-by-step thinking).
DO NOT use generic marketing jargon or corporate speak (Rule #6: What not to do). Write in a direct, practical tone like you’re explaining this to a friend who runs an online store.
Context: This will be published on a blog read by small e-commerce business owners who handle their own marketing and are looking for tactics they can implement this week, not complex strategies requiring enterprise tools (Rule #9: Context).”
Can you feel the difference? The second prompt sets up the AI for success. It knows exactly what to create, who it’s for, what style to use, and what constraints to work within.
The Bottom Line
Writing good AI prompts isn’t about being clever or knowing secret tricks. It’s about being clear, specific, and thoughtful in how you communicate what you need.
Think of AI as a highly capable partner who wants to help but needs good direction. When you master these 9 rules, you’ll stop getting generic, meh outputs and start getting content that’s actually useful.
And here’s the best part: these skills get better with practice. Every prompt you write teaches you something about what works and what doesn’t. You’ll start to develop an intuition for prompting, and what initially took you 10 minutes to craft will eventually take 30 seconds.
Let me be real with you for a second: you’re going to write some bad prompts. We all do. You’ll get outputs that make you think “well, that was useless.” That’s completely normal and actually part of the learning process. The difference between someone who gets frustrated and gives up versus someone who becomes skilled at prompting is simple—the second person treats those “bad” results as feedback.
When you get a mediocre output, ask yourself:
- Was I specific enough?
- Did I provide enough context?
- Did I tell it what NOT to do?
- Should I have broken this into smaller steps?
- Do I need to iterate and refine?
More often than not, you’ll spot the issue immediately. “Oh, I told it to write an article but didn’t specify the length or audience.” “Ah, I forgot to mention the tone I wanted.” “Right, I should have given it examples of what I’m looking for.”
Your Prompting Journey Starts Now
Here’s my challenge to you: pick one of these rules and focus on it for your next five AI interactions. Just one. Don’t try to implement all nine at once (see, I’m taking my own advice about breaking things into smaller steps).
Maybe you focus on being more specific this week. Or you experiment with giving the AI different roles to play. Or you practice the art of iteration instead of expecting perfection on the first try.
As you get comfortable with one rule, add another. Before you know it, writing effective prompts will become second nature.
And remember: AI technology is evolving rapidly. These core principles work now and will likely continue working as models improve, but stay curious and keep experimenting. The best prompters are always learning, always trying new approaches, always pushing to see what’s possible.
The AI revolution isn’t just about having access to these powerful tools—it’s about knowing how to use them effectively. And now you do.
So go forth and prompt better. Your AI awaits your instructions—make them count.
What’s your biggest challenge with AI prompting? Which of these rules are you most excited to try? The beauty of AI is that it’s always ready for another attempt, so there’s no risk in experimenting. Just remember: specific beats vague, context beats assumptions, and iteration beats perfection on the first try.
Now go create something amazing.
Quick Reference: The 9 Rules
Before you go, here’s a handy cheat sheet you can reference:
- Be Ridiculously Specific - Include details about length, audience, format, style, and purpose
- Give the AI a Role - Assign a persona or perspective to activate specific knowledge patterns
- Provide Examples - Show what you want, don’t just describe it
- Break Complex Tasks Down - Tackle big projects step-by-step instead of all at once
- Use Constraints Strategically - Limitations often lead to better, more creative results
- Tell It What NOT to Do - Explicitly state what to avoid to prevent common AI patterns
- Ask for Step-by-Step Thinking - Request reasoning to get more thorough, thoughtful responses
- Iterate and Refine - Treat prompting as a conversation, not a one-shot command
- Provide Context - Give background information so the AI understands your specific situation
Print this out, bookmark it, tattoo it on your arm—whatever helps you remember when you’re staring at that blank prompt box wondering what to write.
The difference between frustration and success with AI often comes down to these fundamentals. Master them, and you’ll wonder how you ever managed without AI as your creative and intellectual partner.