A few months ago, I found myself staring at a blinking cursor, trying to get an AI to help me draft a letter for a friend who was going through a tough time. I typed, retyped, and deleted. Over and over. Nothing sounded right. The responses I got were either robotic or way too vague.
It wasn’t until I shifted how I was asking — not just what I was asking — that the responses started to feel right.
Turns out, prompting AI isn’t that different from talking to humans. The better you understand how to guide the conversation, the more meaningful it becomes. And while there’s no magic formula, there are some habits—simple, often overlooked ones—that seem to separate the “meh” results from the moments where the AI actually feels helpful.
We now live in a world where tools like ChatGPT, Claude, Gemini, and others are becoming our collaborators. But here’s what most people don’t realize: the true power of these tools lies in the question, not just the technology. And according to experts and prompt engineers, there are seven surprising habits the best prompt writers tend to use—often without even realizing it.
Let’s walk through them together. Not in a “Here’s how to be perfect” kind of way. But more like, “Here’s what actually works when you’re tired of AI giving you fluff.”
1. They set the scene before they ask the question.
Think about the last time you asked a friend for advice. You probably didn’t just blurt out “What should I do?” and expect a helpful answer. You gave them context. Maybe you said, “Imagine you’re me. You’ve just finished a project that completely drained you…”
That’s what good prompt writers do. They don’t just give the AI a question—they give it a role to play.
Experts call this “role prompting,” and it’s one of the most powerful habits out there. According to Ethan Mollick, professor at Wharton and AI researcher, assigning a clear identity to the AI—like asking it to act as a therapist, editor, or marketing strategist—can dramatically improve the quality of the output.
When you prompt with “You are a seasoned product manager at a startup,” you’re priming the AI with a mental model. You’re giving it a lens to look through. And from there, its answers become more focused, more human.
Try this:
Instead of: “Give me feedback on this paragraph.”
Say: “Act like a New York Times editor helping a first-time op-ed writer polish this paragraph.”
2. They break big asks into small steps.
This one’s subtle, but crucial.
If you ask AI to write you a 2000-word article, fix your résumé, and make it sound “confident but not arrogant” in one prompt—you’re probably not going to love the result. It’s like asking someone to cook dinner, do your taxes, and tell you why your last relationship failed—all in one breath.
The best prompt engineers use something called chain-of-thought prompting. They guide the AI one step at a time. They might first say, “List 5 angles I could take for this article,” then, “Help me expand on point three with examples.”
In fact, a 2022 study from Google Research showed that when AI models are prompted to reason step-by-step, their accuracy on complex tasks improves drastically.
It’s not just a tech trick. It’s a mindset. Start with curiosity. Then follow the thread.
A bad prompt will be: “Write a business plan for a coffee shop.”
Instead, Say: “First, list the core components of a business plan. Then, for each, I’ll guide you.”
3. They use examples like breadcrumbs.
One of the quickest ways to get better AI responses is to stop asking it to guess what you mean.
If you want it to write something “in your tone,” give it a sample paragraph. If you need a story that sounds “honest but funny,” show it a few lines that hit that vibe.
This is something people often skip because it feels tedious. But the truth is, AI is much better at imitation than improvisation. It learns through patterns. The clearer the pattern, the better the output.
The best way to get good writing out of AI is to prime it with good writing. That doesn’t mean you have to be Shakespeare. It just means showing instead of telling. This can be applied this way; instead of prompting:
“Make it sound more like me.”
Say: “Here’s a paragraph I wrote last week. Use that tone to rewrite the following section.”
4. They give the AI boundaries, not just freedom.
Freedom sounds like a gift, but too much of it can actually be paralyzing. The same applies to AI.
People who write effective prompts know that constraints are not the enemy—they’re structure. They might say, “Give me 3 bullet points,” or “Make it sound like it was written by someone explaining this to a curious teenager.”
Get this point straight: clear constraints improve coherence and relevance. It’s like asking a designer to “just make it look nice” versus saying, “I need a clean look with white space and soft colors.”
We often think clarity kills creativity. But actually, clarity guides it.
A good prompt to this effect could be: “Explain blockchain in under 150 words, using metaphors, for someone who knows how email works but not crypto.” Instead of just making a bland statement like this: “explain blockchain”
5. They treat prompts as drafts, not final products.
This might be the most overlooked habit: they revise. A lot.
Most people give up after one weird or vague AI response. But effective prompt writers iterate. They tweak a sentence, add a constraint, ask follow-ups. In fact, this whole process feels less like dictation and more like a back-and-forth conversation.
You don’t prompt once. You sculpt. Each prompt is a chisel stroke. That hit home for me.
So much of effective prompting is really about curiosity and persistence. You’re not demanding a perfect result—you’re exploring what’s possible.
If you are not satisfied with a response, Instead of saying: “That answer isn’t helpful.”
Say: “That’s not quite what I meant. Could you rephrase it using more concrete examples?”
6. They prompt the AI to show its work.
Here’s something that blew my mind when I learned it: if you ask the AI to “think step by step,” it will actually reason better.
This simple phrasing improves accuracy on logic problems by encouraging the AI model to pause, analyze, and build toward a conclusion. This process can explain why reasoning models excel more on logical tasks.
But this isn’t just for logic puzzles. Even for writing or decision-making, prompting AI to lay out its process can lead to clearer, more grounded responses.
And let’s be honest: sometimes we don’t just want an answer—we want to understand how it got there.
An example prompt to explain this would be: “Think step by step: First, what factors should influence my pricing? Then, apply them to my course.”
Instead of: “What’s the best pricing strategy for my online course?”
7. They write like they’re talking to a smart intern—not a mind reader.
This one’s emotional, not technical.
A lot of people approach AI like it’s some all-knowing oracle. But when you do that, you put way too much pressure on it—and yourself. You expect it to read your mind, decode your goals, and solve all your problems.
But the best prompt writers? They treat AI more like a curious intern. Smart, capable, but not psychic. That small mindset shift makes a huge difference.
You go from frustrated (“Why can’t it get this right?”) to engaged (“How can I guide this better?”). And that curiosity opens the door to more useful results.
“AI is not magic. It’s a mirror. The clearer your instructions, the clearer the reflection.” Let this line stick with you.
This Isn’t Just About AI — It’s About Self-Awareness
There’s something kind of beautiful about how this all works.
Because the more specific you are with AI, the more specific you have to be with yourself. What am I really trying to say? What tone do I actually want? Who is this for?
Crafting a good prompt is, in many ways, a practice in emotional clarity. You’re learning to articulate your needs more clearly—not just to a machine, but to yourself.
And like with any relationship, the quality of the conversation depends on how well you understand what you’re bringing into it.
So if you’ve felt like you’re not “doing it right” with AI, give yourself a little grace. You’re not behind. You’re just learning a new language. And like all languages, the best way to get fluent is to keep playing, keep asking, and keep tweaking the way you say things until they start feeling like you.
If anything, that’s the most human part of all this.