You’ve opened ChatGPT (or your gen-AI tool of choice).
Typed in a request.
Waited a beat…
And what came back?
Vague. Bland. Kind of useless. Obviously AI generated slop.
You’re not alone. Most business owners and operators are just guessing when it comes to prompting. And that’s the problem.
Prompting isn’t a magic trick.
It’s a skill, and most of us haven’t been taught how to do it. Until now.
In this guide, you’ll learn a simple, repeatable way to prompt AI like a pro, so you can stop wasting time, start getting strategic output, and feel confident teaching your team how to use AI without breaking things.
Let’s see if we can fix your prompts in the next 10 minutes.
What’s a Prompt, Really?
At its core, a prompt is just the message you give an AI tool to get a response.
Most people miss one important thing: a prompt isn’t a question (even if you phrase it as one). It’s a set of instructions. And like any set of instructions, how clear, detailed, and structured it is determines the quality of what you get back.
Think of prompting like briefing a new team member on their first day.
You wouldn’t just walk up to them and say, “Hey, write a blog post,” then walk away expecting brilliance.
You’d sit them down and explain what it’s about.
You’d give them the context: who the audience is, why the post matters and what tone to use.
You’d probably say, “You’re writing this in our brand voice, which is warm, clear, and a little cheeky.”
You’d give them an outline, examples of past posts, and maybe even a few things to avoid.
That’s what a good prompt does for the AI.
It says: “Here’s who you are, here’s what we’re trying to do, here’s how I want it delivered.”
When you leave that stuff out, it’s like asking your intern to write your sales page with no background and then getting annoyed when it sounds like ChatGPT wrote it.
That’s exactly how you need to think about prompting.
Why do bad prompts happen to good people?
Because tools like ChatGPT seem smart enough to fill in the blanks.
And they kind of are… until they’re not.
The truth is: AI output is only as good as your input prompt.
If your instructions are vague, generic, or missing key context, the AI will guess. And usually get it wrong.
What makes a good prompt?
Before we dive into prompt frameworks, here’s a quick overview of what separates a strong prompt from a crappy one.
A good prompt typically includes:
- Who the AI is acting as (the role or voice)
- What the task is (clear, specific outcome)
- Any helpful context (audience, goal, tone, format)
- Optional examples (to guide the style or structure)
A quick before-and-after example: Crappy prompt: Write a social media post Better prompt: You’re a brand strategist writing a LinkedIn post to attract small business owners. Highlight the benefits of using AI to save time, and include a strong call to action.
Same task. Totally different quality of output.
Why this matters for your business
Bad prompts don’t only waste time. They:
- give you generic content that almost always needs rewriting
- make it hard to scale AI across your team
- lead to the false belief that AI “just doesn’t work for you”
Prompting is a skill. One that can unlock serious leverage in your marketing, operations, and content workflows (once you know how to use it properly).
And you don’t need to be a tech wizard to get it right!
You just need a clear way to structure what you’re asking for.
Next, I’ll show you the main framework I use to write high-quality prompts that actually work.
The CRIT Framework: How to Write AI Prompts That Actually Work
The reason most AI prompts flop? They’re too vague, too nice, or too lazy.
CRIT fixes that.
This simple framework helps you give AI the right ingredients so it can deliver something useful, strategic, and actually worth using. Think of it like briefing your best team member, not shouting instructions into the void.
Here’s how it works:
C: Context
Set the scene. What’s happening? What does the AI need to know to make this output actually make sense?
Don’t just say “Write an ad.” Say who it’s for, what you’re selling, what problem it solves, and why it matters.
**Example: **“I’m launching a $12 digital budgeting spreadsheet for single mums who feel overwhelmed by bills. It includes a free tutorial video. Write a Facebook ad that explains the value and drives clicks.”
Now the AI knows:
- What you’re offering
- Who it’s for
- What problem it solves
- What you want people to do
- That’s fuel in the tank. No more generic junk.
R: Role
Who should the AI act like? Give it a persona or perspective to write from. By setting a strong role you’re pre-loading it with information.
Examples “Act like a friendly but strategic client success manager who knows how to onboard new clients with warmth and clarity.”
“You’re a top Etsy seller who understands impulse shoppers and how to pitch low-cost, high-desire digital products.”
This step can change the entire quality of the output. The right role = the right mindset = the right message.
I: Instructions
This is where you specify what you want the AI to do.
Start with a strong, clear verb: write, summarise, create, plan, rewrite, outline. Don’t be vague. Don’t ask for “help.” You’re here to delegate, not dilly-dally.
Crappy prompt: *“Help me describe my product.” * Better Prompt: “Write a product description that highlights the features of this $29 Canva template bundle, which I have attached to this prompt. Focus on saving time and making Etsy purchasers feel confident in the product.”
Include important details like product type, audience, use case, and benefits. Lazy inputs = lazy results.
T: Task
Tell the AI what the finished result should look like. What format do you want? Bullet points? Email? Sales page? Something short? Something sassy?
Structure gives clarity. And clarity converts.
Examples: “Format this as a three-paragraph email with a subject line, preview text, and CTA.” “Give me five punchy Instagram captions under 150 characters.” “Write two headline options (one bold, one playful) for A/B testing.”
This step cuts the back-and-forth and gives you outputs you can actually use.
Let’s put it all together…
CRIT Framework Prompting Examples
Let’s say you want ChatGPT to write a welcome email for your new clients. You’ve got a general idea, but the results you’ve been getting? Kind of bland.
Here are three examples
**Prompt 1: **“Write a welcome email.”
AI Output:
Hi,
Welcome to our service. We’re excited to work with you.
Let us know if you have any questions.
Thanks,
The Team
What’s wrong with this response?
It’s generic AF. It has zero personality or relevance to your business. You’ll spend more time rewriting it than if you’d written it yourself.
**Prompt 2: **“Write a welcome email for new clients of a marketing agency. It should thank them, tell them what to expect next, and include contact info.”
AI Output:
Subject: Welcome to [Your Agency Name]
Hi [Client Name],
Thank you for choosing [Your Agency Name]. We’re excited to have you onboard.
Here’s what happens next:
- We’ll send your onboarding form
- Your kickoff call will be scheduled
- You’ll receive access to your client dashboard
*If you have any questions, feel free to contact us. *
Cheers, Your Name
Better. Here’s the difference:
It’s more useful to your new client. It includes structure and clear steps. However, it’s still a bit bland and “template-y.”
Prompt 3: The CRIT Version
You’re assisting a small marketing agency in onboarding new clients smoothly and professionally. The agency values clear communication, great client experience, and a warm, strategic tone. The email comes from the founder but should feel like it’s written by a friendly, helpful client success manager. (This is the context) Act as a client success manager who is friendly, strategic, and highly organised. You write in a clear, conversational tone that helps clients feel confident, welcomed, and guided through their next steps. (This is the role) Write a welcome email for new clients who’ve just signed on. The goal is to thank them, outline the next three steps in the onboarding process, and guide them to book their kickoff call. (These are the instructions) Write the full email body (under 250 words). Include a warm thank-you for signing on. Use bullet points to clearly outline the next 3 steps: Complete the onboarding form Schedule the kickoff call Get access to their project dashboard End with a friendly call to action prompting them to book the kickoff call. Sign off as the agency founder. Keep it conversational, professional, and easy to follow. (This is the task)
AI Output:
Subject: Welcome! Here’s what’s next.
Hi [Client Name],
We’re so excited to have you on board. Thank you again for choosing us! This next chapter should feel smooth, smart, and seriously productive, and we’re here to make that happen.
Here’s what happens next:
- Fill out your onboarding form You’ll receive a short onboarding form to help us tailor your setup and come prepared. [Insert link]
- Book your kickoff call We’ll walk through your goals and map out your first steps together. [Insert link]
- Get access to your client dashboard Once your form’s in, we’ll set you up with your project hub: files, timelines, notes, all in one place.
Ready to dive in? Book your kickoff call here: [link]
Talk soon, [Your Name]
Let’s break it down:
It’s warm, personal, and specific It covers exactly what the client needs to know (and nothing more) It saves you from multiple rewrites. This could go out as-is or with minor quick edits
CRIT Framework: In a Nutshell
That’s what CRIT stands for, and it’s what separates pro-level outputs from generic, low-value sludge.
If you’re getting vague or off-target results, chances are one (or more) of these is missing.
Want Even Better Results? Try Iterative Refinement (Your New AI Superpower)
Most people treat prompting like a one-shot deal.
They write something, get a result, and if it’s not perfect… they delete it and start over.
That’s not how the pros do it.
If you want consistently great results, you need to treat AI like a creative partner. Not a vending machine.
**What is iterative refinement? **It’s just a fancy way of saying:
Test. Tweak. Improve.
Instead of tossing out a prompt and hoping for the best, you build and polish it over a few quick rounds. You review what the AI gave you, give it clear feedback and suggestions, and watch it improve.
The simple 4-step method to Iterative Refinement
Step 1: Start with a clear, structured prompt (like CRIT). Step 2: Run the prompt and review the output. Step 3: Ask yourself “is this accurate, helpful, and on-tone? What’s missing or not quite right?” Step 4: Refine your prompt or give follow-up instructions to improve the output.
**Example: Refining the welcome email **Let’s say you liked 80% of the AI’s draft, but it felt too stiff and formal.
Try this follow-up: “Rewrite this in a warmer, more human tone. Use contractions, keep it casual, and make it sound like a real person wrote it.”
Iterative Refinement: In a Nutshell
Don’t scrap a prompt if it flops, refine it.
Iterative refinement means treating AI like a creative partner, not a vending machine.
Test the output, tweak your instructions, and guide it step by step until the result hits the mark.
It’s faster (and smarter) than starting over every time.
Next-Level Prompting: Teach the AI to Think with Chain-of-Thought
Once you’ve nailed clear prompts and learned to refine them, you’re ready for a pro move: chain-of-thought prompting.
What is chain-of-thought prompting? Instead of asking for an instant result, you prompt the AI to think step by step before answering.
Try saying: “Don’t just give me the answer. Show me your thinking.”
This makes the AI more accurate, more logical, and often more creative.
**How to use it **Add phrases like: “Let’s work this out step by step.” “Break this into three phases.” “Start by asking me questions to clarify, one at a time.”
This can really help with content strategy, planning, and decision support.
Chain of Thought Prompting: In a Nutshell
Chain of Thought prompting gets the model to reason step by step, instead of rushing to a surface-level response.
Use it when accuracy, logic, or nuance matters - especially for planning, strategy, or complex decisions.
Use the Layered Prompt Method when tackling bigger projects
If you’re working on something bigger, like planning a launch, building a course, or writing a whole website the CRIT framework can help, but you will almost certainly need more than one prompt.
This is where a technique called layered prompting comes in. It’s a multi-step process that helps you break complex projects into logical, focused pieces the AI can handle one phase at a time.
Here’s how it works
**Step 1: Planning Prompt **Start with the end in mind. Don’t ask for content yet. Ask for the roadmap.
“Step 1. The outcome I’m aiming for is [insert goal]. Before we jump into execution, help me break it down into clear stages. What needs to happen first, what follows, and what’s dependent on what?”
What you are looking for here are
- Major phases
- Deliverables for each
- Dependencies or blockers
Step 2: Deep Dive Prompts Once you have your roadmap, zoom in.
“Step 2. Focus only on [Phase One] for now. Based on the structure we just created, here’s what I need from you: [insert instructions].”
You’re giving the AI a narrowed scope and previous context which massively improves the quality of its work.
**Step 3: Refinement Prompt **Then sharpen the result.
“Step 3. Using your outline and deliverables from step 1, and your instructions from step 2, refine what you just wrote. Cut what’s vague, expand what’s useful, and rebuild anything that doesn’t hit the mark. Deliver the stronger version.”
This step turns decent into excellent.
**Step 4: Integration Prompt **Once each phase is refined, pull everything together into a final deliverable.
“Combine all previous outputs into one cohesive deliverable. It should read smoothly, feel unified, and meet the original brief without repeating or conflicting details. This version should be aligned with our brand voice and ready to use.”
Remember to thread your context by referring back to previous prompts. This keeps the AI grounded and focused and saves you from starting over every time it forgets the brief.
Layered Prompting: In a Nutshell
Layered prompting breaks complex requests into manageable steps, each building on the last.
Plan the roadmap, zoom into one phase, refine the results, and only then stitch everything together.
It saves time, avoids overwhelm, and gives you outputs you can actually use.
Final tip before you go
No matter which prompting method you use - CRIT, Chain of Thought, or Layered - once you’ve landed on an output you’re actually happy with, don’t stop there.
Ask your AI:
“This output is exactly what I was aiming for. What could I have asked as my initial prompt to get this result sooner?”
It’s one of the simplest ways to reverse-engineer better prompts, improve your inputs over time, and teach your team to do the same.
Create a prompt library that you share with your team so these can be re-used to save everyone time, effort and ensure more consistent results.
Better prompts are one piece. Calm systems and a citeable voice are the rest. Related reading: