How to Get Better Results from ChatGPT: A Plain-Language Prompting Guide

ChatGPT gives you bad answers because you’re asking bad questions. Here’s how to ask the right way.

• The 5 principles that turn vague prompts into powerful ones
• The context sandwich method (top slice, middle, bottom slice)
• How to show examples and iterate until you get what you want
• Real before-and-after examples of prompts that work
• Common mistakes that kill your results

You probably already know this: ChatGPT can feel like talking to someone who understood about half of what you said.

You ask it for a social media post and get something generic. You request email copy and it comes back sounding like a robot. You're thinking, "Maybe AI isn't for me." But here's the truth—it's not that AI is bad. It's that we're not asking it the right way.

I used to get frustrated with ChatGPT all the time. Then I realized I was being vague. I was asking it things like "Write me a post" or "Create an email" and wondering why the results felt flat. Once I started being specific about what I actually wanted, everything changed.

This guide walks you through the exact techniques that turned ChatGPT from a disappointing tool into something that actually saves me hours every week.

Why AI Gives You Bad Answers (And It's Usually Your Fault)

Let's be honest: when ChatGPT gives you garbage, it's usually because you asked a garbage question.

This isn't about the AI being dumb. ChatGPT is reading what you type and trying to guess what you mean. If you give it vague instructions, it has to make assumptions. And those assumptions often miss the mark.

When you tell ChatGPT "Write a product description," it doesn't know:

• What your product actually does
• Who's buying it
• What language your customers use
• How long you want it
• What tone fits your brand
• What problems it solves

So it writes a generic description that could fit any product. Then you read it and think the AI is useless.

The secret is: you have to do the thinking for the AI, not the other way around. You tell it exactly what you need, show it examples, give it context. Then it works.

The Five Principles That Actually Work

1. Be Specific (Not Vague)

"Write a social media post" = bad.
"Write an Instagram caption for a new workout program, 100-150 words, written for Black women ages 40+, funny tone, should make them feel capable not intimidated" = good.

When you get specific, the AI knows what lane to stay in. It can't wander.

Before: "Create a subject line for an email."
After: "Create an email subject line for a business coach reaching out to women who have never hired a coach before. They might be skeptical about wasting money. Make it curious but honest, under 50 characters."

See the difference? The second one gives ChatGPT actual instructions.

2. Give It Context (The Sandwich Method)

This is my favorite technique because it's simple and it actually works. I call it the "context sandwich."

You're putting information around your request, like bread around the filling.

Here's how it works:

Top slice (who is this for?): Tell ChatGPT who your audience is. Age range, what they worry about, what they know and don't know, why they might hesitate.

Middle (your actual request): The specific thing you want it to write or do.

Bottom slice (what's the goal?): Tell it what you want to happen after someone reads this. Should they click? Should they feel less scared? Should they understand something new?

Example:

Top slice: "I'm writing for women ages 35-55 who have always worked for someone else. They're thinking about starting a small business but they're nervous because they have bills to pay and can't afford to fail."

Middle: "Write a short email (about 150 words) telling them about a free workshop on creating a business plan."

Bottom slice: "The goal is to get them to sign up even though they're scared. So make them feel like it's safe to try and that other people like them have done this."

That sandwich gives ChatGPT everything it needs. It's not guessing anymore.

3. Show It Examples

If you have an email, social post, or piece of writing you actually like, show it to ChatGPT.

You can paste it right into the chat and say: "Write something like this, but about [your topic]."

ChatGPT will study the tone, length, structure—all of it. And it'll apply that style to whatever you're asking for.

Example:

"Here's an Instagram post I wrote that got good engagement:

[paste your post here]

Now write a similar post about my new digital course. Keep the same tone and length, but update it for the course."

This is like showing someone a photo of what you want instead of describing it. Way more effective.

4. Ask for the Format You Actually Want

Don't just ask for "copy." Tell ChatGPT the exact format.

"Write a 3-email sequence" is better than "write emails."
"Create a list of 10 subject lines, formatted as a numbered list" is better than "give me subject lines."
"Write a 400-word blog post with an intro paragraph, two main sections with subheadings, and a conclusion" is better than "write a blog post."

The more specific about format, the less editing you do on the back end.

5. Iterate (Ask Follow-Up Questions)

This is the part people skip and it's a mistake.

One prompt doesn't always get you the final product. That's normal and it's okay.

If ChatGPT gives you something close but not quite right, don't start over. Just tell it what to adjust:

• "Make it shorter"
• "Make it more funny and less formal"
• "Add a question at the beginning that hooks the reader"
• "Remove the part about pricing and focus on the benefits"

Each time you ask, it gets closer to what you actually want.

Real-World Examples: Before and After

Here's what I mean when I say specificity changes everything.

Example 1: Social Media Post

Bad prompt (what most people do):
"Write a Facebook post about my coaching program."

ChatGPT's response:
"Excited to announce my coaching program! I help women achieve their goals and live their best lives. My program includes personalized coaching, resources, and community support. If you're ready to transform your life, click the link below."

(Ugh, right? Generic and a little cheesy.)

Good prompt (context sandwich style):
"I'm writing for women ages 45-60 who have always put everyone else first—kids, spouses, jobs. They're curious about personal growth but they think that's selfish. They don't have a lot of disposable income and they're skeptical of coaching. Write a Facebook post (2-3 short paragraphs) that acknowledges their hesitation, tells them that taking care of themselves isn't selfish, and invites them to a free intro call. The tone should be warm, honest, and like a friend talking to them."

ChatGPT's response:
"Here's a question I ask myself: When did taking care of yourself become selfish? Because I'm watching women in my community put everyone else's needs first, and then wonder why they're exhausted. And when I mention coaching, they get quiet. Like investing in yourself is somehow wrong. It's not. You taking an hour to figure out what you actually want? That's not selfish. That's smart. That's important. If this feels true for you, I'd love to talk. I'm offering free intro calls for women who are curious but not sure if coaching is for them. No pressure. Just a real conversation. Let's see if we're a good fit."

(See? This one actually sounds like a real person. It addresses the specific hesitation. It works.)

Example 2: Email Subject Line

Bad prompt:
"Give me email subject lines."

ChatGPT's result:

• "Check Out Our Latest Offer"
• "Don't Miss Out Today"
• "Exclusive Deal Just for You"

(You've seen these a thousand times. They don't stand out.)

Good prompt:
"I'm sending an email to women who have been thinking about starting a business but haven't taken the first step. Some of them are worried they're too old, some think they need a huge amount of money upfront, some just don't know where to start. I want to send them a free resource called 'Your First 30 Days as a Business Owner.' Create 5 subject lines that speak to a specific fear (age, money, not knowing where to start), acknowledge that fear is normal, and make them curious about the resource. Keep each under 50 characters."

ChatGPT's result:

• "Not sure where to start? I'll show you."
• "Too old to start a business? Absolutely not."
• "Starting with $0? Here's how real women did it."
• "30 days to your first business decision"
• "The one thing nobody told me before starting"

(Now these feel real. Each one speaks to something someone is actually worried about.)

Common Mistakes to Avoid

Mistake 1: Asking for too much at once.
Don't ask ChatGPT to write your entire marketing strategy in one prompt. Break it into pieces. Ask for one social post, then another, then an email.

Mistake 2: Not being honest about what you don't know.
If you're not sure about your audience, say so. If you don't have examples, that's fine too. ChatGPT can help you figure those things out. Just be upfront about where you're starting from.

Mistake 3: Accepting the first response without feedback.
The first draft is rarely the final product. That's true whether you write it yourself or ChatGPT writes it. Always ask for adjustments.

Mistake 4: Not giving background information.
ChatGPT can't read your mind. It doesn't know your business, your customers, or your brand. You have to tell it.

Mistake 5: Forgetting to ask for tone.
Tone changes everything. Always specify: funny or serious? Warm or professional? Casual or formal?

Try It With AI

Ready to put this into action? Copy any of the prompts below, paste it into ChatGPT or Claude, fill in the [BRACKETS] with your info, and hit send. You'll have a solid first draft in minutes.

[TOP SLICE] I’m writing for [YOUR IDEAL CUSTOMER description]. [MIDDLE] Write me [WHAT YOU WANT: email/post/headline]. [BOTTOM SLICE] The goal is to [WHAT YOU WANT TO HAPPEN: get signups/build trust/make them laugh]. Make it [TONE: warm/funny/direct/professional]. [LENGTH/FORMAT SPECS].

Using the context above, write me [NUMBER] versions of [WHAT YOU WANT]. Each version should: [SPECIFIC INSTRUCTIONS FOR HOW THEY DIFFER]. I’ll be editing and choosing between these, so give me real variety to work with.

I liked the [WHICH ONE] version best. But can you adjust it to be [SPECIFIC FEEDBACK: shorter/funnier/more specific/less pushy]? Keep the core idea the same but make this one change: [YOUR REQUESTED CHANGE].

ChatGPT isn’t bad at writing. You were just asking it to guess what you meant. Be specific and it becomes a tool that saves you hours every week.