Talking to AI Without Losing Your Cool: How to Write Prompts That Actually Work

"Learn how to write AI prompts that actually work. Avoid vague questions, get smarter answers, and never let AI hand you the wrong cake again."

A Norwegian man, Lars Henriksen (name changed), was simply sitting at home one afternoon. Out of curiosity, he decided to test ChatGPT. He wondered: what does the AI actually know about me?

So he typed the simplest question he could think of:

“Who is Lars Henriksen?”

Nothing complicated. No long details. Just his name.

What happened next left him stunned. Instead of a polite “Sorry, I don’t know,” or maybe a short, harmless line about his background, the AI replied with something straight out of a crime thriller.

It confidently described Lars as a man who had murdered two of his children, gone through a trial, and was now serving a 21-year prison sentence.

But here’s the truth: none of it was real.

Lars had never committed a crime in his life. Yet the AI, with complete confidence, invented a story that turned him into a monster.

Imagine the shock—sitting in your living room, coffee in hand, and suddenly discovering that an AI has cast you as the villain in a murder drama.

So what happened?

Can AI really make up lies like that?

Was the system broken?

Or was the real problem the way the question was asked?

What Happened and Why Prompts Matter

So, why did Lars Henriksen get accused of crimes he never committed? Let’s unpack it.

Here’s the thing: AI can absolutely provide facts, summaries, and data when the information exists in its training or when connected to reliable sources. That’s what makes it so useful. But when the AI doesn’t have enough reliable data about something—or someone—it doesn’t just shrug and say, “Sorry, I don’t know.”

Nope. Instead, it tries to be “helpful.” It starts filling in the gaps with things that seem likely. It guesses. It improvises. And it delivers those guesses with full confidence, like the overconfident friend who would rather make up an answer than admit they don’t know.

That’s what happened to Lars. He asked a vague, open-ended question:

“Who is Lars Henriksen?”

With no reliable public data to lean on, the AI had two options:

1. Admit it couldn’t find anything.

2. Or… invent something that sounded plausible.

Unfortunately, it chose option two. It stitched together random elements—Norwegian-sounding details, legal terms, and a prison sentence—until it looked like a real story. Suddenly Lars, an innocent man, was written into a made-up crime drama.

It’s the digital equivalent of walking into a bakery and saying: “Give me something to eat.” The baker, eager not to disappoint you, grabs the first thing he sees: a giant birthday cake with “Happy 80th Grandma” written across it. 🎂

Technically, yes, he gave you food. But it wasn’t what you wanted, it wasn’t useful, and it definitely wasn’t meant for you.

That’s the problem with vague prompts. The AI will always try to give you an answer—even if it has to invent one.

And this is the big takeaway: prompts aren’t just casual questions; they’re instructions. The way you phrase them controls what kind of answer you get back.

Vague prompts = vague or unreliable answers.

Clear, specific prompts = sharper, safer, more accurate answers.

In Lars’s case, a better prompt would have saved him from this nightmare. Something like:

> “Provide only verifiable information about Lars Henriksen. If no reliable sources exist, say so clearly. Include sources I can cross-check. Do not invent details.”

See the difference? One line of extra clarity changes the entire outcome.

So if you’ve ever asked AI a question and thought, “That’s not what I meant at all,”—well, now you know why. The AI wasn’t broken. The prompt was.

The Classic Mistakes People Make

By now, you can probably see why Lars’s vague question went so horribly wrong. But let’s be honest—he’s not the only one guilty of bad prompting. Most of us have sat in front of an AI tool, typed something quick, and then wondered why the answer felt boring, random, or just plain wrong.

That’s because writing prompts isn’t about what you think in your head—it’s about what you actually type on the screen. And most of the time, we don’t give AI enough to work with.

Here are the four big mistakes people make when talking to AI (and how to fix them).

1. Being Too Vague

This is mistake number one. Think of it like walking into a library and saying, “Tell me about history.” The librarian has no clue whether you want world history, American history, the history of bread, or the evolution of Justin Bieber’s haircuts.

AI is the same. If your question is too broad, you’ll get a broad, bland answer that doesn’t really help.

Bad Prompt: “Tell me about dogs.”

AI’s Answer: A generic overview about canines, their breeds, and maybe a fun fact about wagging tails.

Better Prompt: “Tell me about bulldogs.”

Why it’s better:

By narrowing the scope, you’ve given AI a smaller sandbox to play in. Instead of rambling about all dogs, it zooms in on bulldogs—their wrinkly faces, funny snorts, and stubborn personalities. Still a general answer, but now it’s focused, useful, and not all over the place.

2. Forgetting Context

This is different from being vague. Here, the topic might be clear, but you forget to tell AI who the audience is and what the purpose is.

Without context, AI writes in a generic, one-size-fits-all voice. With context, it suddenly knows how to tailor the answer.

Bad Prompt: “Write a speech.”

AI’s Answer: Some bland motivational piece that could be slapped on a poster in a dentist’s office.

Better Prompt: “Write a lighthearted wedding toast for my best friend who once tripped over a goat in Spain. Make it funny but not embarrassing.”

Why it’s better:

The topic is still “a speech,” but now AI knows:

who it’s for (wedding guests),

what it’s about (your best friend),

and the tone (funny but kind).

The result is personal, memorable, and way more useful than a generic “ladies and gentlemen” snoozefest.

3. Asking for Too Much at Once

Here’s a biggie. You throw your entire wish list at AI in one breath and hope it spits out perfection. Spoiler: it won’t.

When you overload AI with ten different requests, it tries to juggle them all at once—and you end up with a messy answer that satisfies none of them.

Bad Prompt: “Write me a 10-page essay about global warming, renewable energy, climate policy, and how to save the planet, and make it funny.”

AI’s Answer: A wall of text that’s part school report, part comedy sketch, part lecture—and none of it lands well.

Better Prompt: “Write me a two-paragraph summary of why global warming matters, in a casual, easy-to-read tone.”

Why it’s better:

This version doesn’t overwhelm the AI. You’ve broken the job into one clear step: summarize the importance of global warming in a casual style. Once you’ve got that, you can build further—ask about renewable energy next, then climate policy, then maybe sprinkle in humor. Step by step = cleaner, better answers.

4. Expecting Perfection in One Try

Here’s a mindset mistake. People think of prompts like magic spells: say the words once, and poof—perfect answer.

But that’s not how AI works. It’s a conversation. You start broad, then refine. You push it, redirect it, and shape it into what you want.

Mistake: Typing one vague question, hating the answer, and closing the app.

Better Approach: Treat it like brainstorming. Start with: “Write me a short summary of X.” Then refine: “Make it shorter.” → “Add humor.” → “Rewrite it as a poem.”

Why it’s better:

This back-and-forth lets you gradually polish the answer until it fits perfectly. The real power of AI isn’t in the first draft—it’s in the revisions you guide it through.

Pro-Level Tricks (For When You Want to Sound Like a Prompt Ninja)

Once you’ve mastered the basics of prompting—avoiding vagueness, adding context, and giving constraints—you can step up your game with some pro-level tricks. These aren’t complicated, but they make AI feel less like a clumsy intern and more like a sharp assistant who gets you.

Here’s the toolkit:

1. Role-Playing Prompts

One of the easiest ways to get better answers is to tell the AI who it should pretend to be. This sets the tone, the style, and the level of expertise.

Example: “You are a fitness coach. Create a 7-day workout plan using only dumbbells.”

Result: Instead of vague tips like “exercise regularly,” you’ll get a structured plan with sets, reps, and rest days.

Another Example: “You are a stand-up comedian. Write 5 short jokes about coffee addiction.”

Result: Punchlines that sound like they belong on stage, not in a textbook.

Why it works: Without direction, AI floats between styles. Role-playing forces it to pick a lane.

2. Force Honesty

AI has a bad habit of “making things up” when it doesn’t know. That’s how poor Lars ended up accused of crimes he never committed.

You can stop this with one simple rule in your prompt: “If you don’t know, just say you don’t know.”

Example: “Tell me about the author of this obscure book. If you’re not sure, admit it. Don’t invent details.”

Result: Either you get real, sourced info—or a clear “I don’t know.” No hallucinated drama.

Why it works: You take away the AI’s pressure to always have an answer.

3. Formatting Control

Sometimes the answer itself is fine, but the format is a mess—giant walls of text you don’t want to read. The fix? Tell AI exactly how to present its response.

Example: “List 10 quick morning habits for productivity. Use bullet points with emojis. Keep each point under 8 words.”

Result:

☀️ Wake up at the same time

🥤 Drink a glass of water

📝 Write a 2-minute to-do list

Now it’s easy to read, skimmable, and way more engaging.

Why it works: AI thrives when it knows how to package the answer.

4. Step-by-Step Prompting

Big tasks overwhelm AI. If you ask for a 20-page report in one shot, you’ll probably get something messy and inconsistent. The smarter way is to build in layers.

Step 1: “Make me an outline for a blog about good prompts.”

Outline is created. After that each section is expanded by next prompt.

Step 2: “Expand section 1 into 500 words with a casual tone.”

The created section needs improvement. Specific improvement is added.

Step 3: “Now add a funny analogy to that section.”

Similarly steps 2 and 3 is done for each section. And finally you have a 20 page report. You’re basically guiding AI the way a teacher guides a student—start broad, then fill in details.

Why it works: Breaking things down reduces errors, keeps the content focused, and gives you more control.

5. Add Constraints

Constraints aren’t limitations—they’re magic. They sharpen answers by giving AI boundaries.

Example: “Explain black holes in exactly 100 words, like you’re talking to a 10-year-old.”

Result: Clear, digestible, to the point.

Another Example: “Write a horror story in three sentences.”

Result: A tight little narrative that doesn’t drag.

Why it works: Without constraints, AI rambles. With them, it delivers something crisp.

And remember—these tricks don’t just make AI answers better. They make them yours: tailored, practical, and on point

Final Thoughts

AI is the future, and it’s here to stay. Whether you love it, fear it, or still don’t quite get it, there’s no denying it’s going to be part of our lives from now on. And like any new tool, it comes with a learning curve.

One of the biggest skills you can pick up right now is learning how to prompt. Some people roll their eyes at this and call it “learning to be the questions guy.” But let’s be real — asking the right questions has always been powerful. In school, in jobs, in life… and now, in AI.

The truth is, many people try AI once, get a weird or wrong answer, and then give up. They think, “This thing doesn’t work.” But it wasn’t the AI that failed — it was the prompt. The way you ask is the way you’ll receive.

We hope this article made you a little more educated about how to talk to AI without losing your cool. And if you’ve got doubts, thoughts, or even topics you’d like us to cover in the next post — drop them in the comments.