How to Write Better Prompts for ChatGPT – 7 Proven Techniques

Ever asked ChatGPT a question and gotten back something generic, vague, or just plain unhelpful? You’re not alone. I’ve been there too—typing away, hoping for gold, and getting… well, mush.

The secret isn’t the AI. It’s you. Or more specifically, it’s how you talk to the AI.

Prompt engineering sounds fancy, but it’s really just learning how to ask better questions. Think of it like this: if you walked up to a human expert and mumbled “tell me about stuff,” you’d get a shrug. But if you said, “I’m trying to solve X problem with Y constraints—what’s your approach?” suddenly you’re getting real value.

Let me walk you through 7 techniques that’ll transform your ChatGPT results from meh to magnificent.

1. Be Specific About What You Want

Here’s a before-and-after that changed everything for me:

Before: “Write me a workout plan.”

After: “Create a 4-week beginner workout plan for someone who can only exercise at home, has dumbbells up to 20 lbs, and wants to focus on building strength. Include rest days and explain why each exercise matters.”

See the difference? The second prompt tells ChatGPT exactly what constraints you’re working with, what your goal is, and even asks for reasoning. You’re not just asking for content—you’re asking for the right content.

Common mistake: Being too vague. “Help me with marketing” could mean anything from social media strategy to email copy to SEO. Pick one.

2. Give It a Role to Play

ChatGPT is like a method actor—it performs better when you tell it who to be.

Before: “How do I fix a leaky faucet?”

After: “You’re a licensed plumber with 15 years of experience. Walk me through fixing a dripping kitchen faucet step-by-step, including what tools I’ll need and common mistakes beginners make.”

When you assign a role, you’re not just getting information—you’re getting expert-level information with the right tone and depth. I use this for everything now: “You’re a patient coding tutor,” “You’re a nutritionist who specializes in plant-based diets,” “You’re a career coach for tech workers.”

Try this: Next time you ask a question, start with “Act as a [specific expert]…” and watch the quality jump.

3. Show, Don’t Just Tell (Use Examples)

Want better output? Show ChatGPT what good looks like.

Before: “Write product descriptions for my online store.”

After: “Write product descriptions for my online store. Here’s the style I like:

Example 1: ‘This isn’t just a coffee mug—it’s your morning ritual in ceramic form. The double-wall insulation keeps your brew hot for hours, while the ergonomic handle fits perfectly in your grip whether you’re right or left-handed.’

Example 2: ‘Meet your new favorite jeans. Engineered with 2% elastane for stretch that moves with you, these maintain their shape wash after wash. The mid-rise cut flatters every body type.’

Now write descriptions for: [product list]”

This technique is called “few-shot prompting,” and it’s incredibly powerful. You’re not explaining what you want—you’re demonstrating it.

4. Break Complex Tasks Into Steps

Here’s where most people go wrong: they ask for everything at once.

Before: “Help me start a podcast.”

After: “Let’s work through starting a podcast step by step. First, help me identify my target audience and niche. I’m thinking about [topic]. Ask me questions to clarify my direction before moving to the next step.”

By breaking things down, you get better results at each stage. Plus, ChatGPT can ask you questions, which makes the whole thing more interactive and tailored.

I use this for big projects all the time. Writing a book? Start with the outline. Building an app? Start with the features list. You wouldn’t build a house without a blueprint—don’t ask AI to create something complex without a roadmap.

5. Tell It What NOT to Do

Constraints are your friend. Seriously.

Before: “Write a blog post about time management.”

After: “Write a 1,200-word blog post about time management for remote workers. Don’t use clichés like ‘work smarter, not harder.’ Don’t mention the Pomodoro Technique—I already know about it. Focus on psychological barriers to productivity instead of tools and apps.”

By ruling things out, you’re steering ChatGPT away from the generic advice it defaults to. You’re saying, “I’ve done some research—give me something deeper.”

Pro tip: If you keep getting responses that are too long, add “Keep this under 500 words.” Too formal? Add “Write in a casual, conversational tone like you’re explaining to a friend.”

6. Ask It to Think Before Answering

This one feels weird, but it works surprisingly well.

Before: “What’s the best way to learn Spanish?”

After: “Before answering, think through the different approaches to learning Spanish: self-study apps, tutors, immersion, classes. Consider the pros and cons of each for a busy professional with 30 minutes a day. Then give me your recommendation with reasoning.”

When you ask ChatGPT to reason through something, you get more thoughtful answers. It’s the difference between a gut reaction and a considered opinion.

I use this for decisions especially: “Compare these three options and explain your reasoning before recommending one.”

7. Iterate and Refine

Your first prompt doesn’t have to be perfect. Treat it like a conversation.

Start with: “I want to write a children’s book about a robot who learns emotions.”

Then follow up with: “That’s good, but make the tone more playful. Also, the main character should be 8 years old in robot years—what does that mean?”

Then: “Great! Now outline the first chapter with three key scenes.”

Some of my best results come from the third or fourth message in a conversation, not the first. ChatGPT remembers context, so use that to your advantage.

Common mistake: Giving up after one bad response. If the answer isn’t right, tell it what’s missing and try again.

Putting It All Together

Let me show you a master-level prompt that uses all these techniques:

“You’re an experienced career coach who specializes in helping people transition into tech. I’m 35, work in retail management, and want to become a software developer.

Before giving advice, ask me 3-5 questions about my situation, goals, and constraints.

Once you understand my context, create a 6-month learning plan that accounts for:
– My full-time job (40 hours/week)
– My budget of $200/month for courses
– My goal of landing an entry-level developer role

Don’t suggest bootcamps over $5,000. Don’t tell me to ‘just network more’ without specific strategies.

Format the plan as a week-by-week breakdown with specific resources and milestones.”

See how that combines role-playing, specificity, constraints, examples, and step-by-step thinking? That’s the kind of prompt that gets you results worth acting on.

Common Mistakes to Avoid

Let me wrap up with the biggest errors I see people make:

1. Asking yes/no questions when you want depth. “Is Python good for beginners?” gets you a paragraph. “Compare Python, JavaScript, and Ruby for someone who wants to build web apps—what should I learn first and why?” gets you actionable advice.

2. Not providing context. ChatGPT doesn’t know you’re a beginner, an expert, a student, or a CEO. Tell it.

3. Accepting the first draft. If something feels off, say so. “This is too technical—explain it simpler” or “Give me more specific examples” are both valid follow-ups.

4. Forgetting that you can edit its output. Got a good response that’s too long? “Summarize this in 3 bullet points.” Too dry? “Rewrite this with more personality.”

Advanced Technique: Chain of Thought Prompting

Here’s a pro technique that separates beginners from power users: chain of thought prompting.

Instead of just asking for an answer, ask ChatGPT to show its work:

Basic: “What’s 23 × 47?”

Chain of thought: “Calculate 23 × 47. Show your work step by step, explaining each part of the multiplication.”

This works for complex problems too:

“Plan a 5-day trip to Rome for a family with two teenagers. Think through:
1. What teenagers typically enjoy
2. How to balance sightseeing with downtime
3. Budget considerations for a family of four
4. Weather in [month]
5. Transportation options

Show your reasoning for each day before presenting the final itinerary.”

When you ask AI to think out loud, you get better results—and you can spot flaws in its reasoning.

Real-World Prompt Templates

Let me give you some copy-paste templates you can use immediately:

For learning something new:
“I’m a complete beginner at [topic]. Explain [specific concept] in simple terms with real-world examples. Then give me three practice exercises to test my understanding. Don’t move to advanced topics until I demonstrate I grasp the basics.”

For problem-solving:
“I’m trying to [goal]. I’ve already tried [X] and [Y], but they didn’t work because [reason]. What are three alternative approaches I haven’t considered? For each, explain the pros, cons, and what I’d need to get started.”

For content creation:
“Write a [type of content] about [topic] for [audience]. Tone should be [adjective]. Include [number] specific examples. Avoid [things to avoid]. Keep it under [word count] words. Start with a hook that [desired effect].”

For decision-making:
“I’m deciding between [option A] and [option B]. My priorities are [list]. My constraints are [list]. Compare these options across my priorities, then recommend one with clear reasoning. Also tell me what information I’m missing that would help decide.”

Save these templates. Customize them. They’ll save you hours.

Your Turn

Here’s what I want you to do: take a prompt you’ve used before that gave you disappointing results. Rewrite it using at least three of these techniques. Then try it.

I promise you’ll be shocked at the difference.

The truth is, ChatGPT is incredibly capable—but it needs direction. You’re the director. The better your instructions, the better the performance.

So go ahead. Be specific. Give it a role. Show examples. Break things down. Set constraints. Ask it to think. And iterate until you get exactly what you need.

That’s how you write better prompts. That’s how you get better results.

What will you ask it first?