I’ve Spent 104 Hours Writing ChatGPT Prompts. Here’s What I Learned

By Matt Giaro

✦ FREE EMAIL COURSE: 5 Secrets to create meaningful (and personal) content 10X faster with AI

I was dead wrong about ChatGPT.

I first thought that it was for lazy chumps who wanted to get rich with just a click of a button.

But as the saying goes, “Only stupid people never change their minds.”

Over the past few weeks, I went all in and geeked out on deciphering how to use Sam’s new tool to improve my writing workflow.

Here’s what I learned.

How to craft unmatched (and damn awesome) AI output your competitors would dream of

“AI content is dry and boring.”

Ever heard this platitude? Well, I was one of those yelling trolls.

Slapping together a half-hearted prompt and expecting Shakespeare in return is like feeding a gourmet chef dollar-store ingredients and expecting a Michelin-starred meal.

I used many of those stupid prompts, like:

  • Write 10 headlines about X
  • Write an outline for my next Y blog post.
  • Brainstorm 10 ideas for a viral article about Z


They’re so damn vague and generic that I can smell AI-generated stuff from miles away.

The reason? The input is garbage.

In case you didn’t notice, ChatGPT can’t read your mind (yet). Prompts are the raw material from which the beast cranks out its content.

Do this right, and you can avoid the pit of despair that comes with reading content that misses the mark by a mile. Do this wrong, and you’ll end up competing against the schmucks.

When you hand over a solid, unambiguous prompt, it becomes a launchpad for AI to take flight.

Say you’re writing a tech blog.

Instead of saying, “Write about the latest tech trends,” say, “Discuss the impact of quantum computing on data security in the cloud.”

But that’s only the first step.

Stop being lazy

This brings me to the next point.

Crafting powerful prompts is like writing a blog post: they require a lot of typing. In fact, I like to say that the best prompts I’ve engineered are the size of blog posts.

Here’s what I mean.

By fleshing out your prompt as long as a blog post, you’ll get a coherent and well-structured piece in which you mention what exactly you want in its finest details:

  • What tone should it use?
  • For which audience is it for?
  • What structure should it follow?
  • Who are you to talk about the topic?
  • etc.

And yes, those blog posts will need some revisions until you fine-tune them to your own taste.

Great prompts aren’t one-liners. They are blog posts.

Create your clone

Handing AI a crumpled map and expecting it to find treasure is wishful thinking.

There was this time when I gave AI a vague prompt with no real structure, tone, or personality to emulate.

The result? A jumbled mess that read like a mismatched puzzle, leaving me feeling like I’d fallen flat on my face.

Here’s the big secret: having a successful model for your AI to emulate is not an option; it’s a necessity. Do this, so you can bypass the gut-wrenching frustration of picking through a chaotic pile of words.

Let me break it down for you.

If you want your AI to sound like Hemingway, feed it Hemingway.

If you want it to write an engaging how-to guide, give it an example of a how-to guide that you admire. The more specific and accurate your model is, the closer your AI will get to hitting the bullseye.

Don’t just wing it. Find a stellar model for your AI to follow.

Decrypting AI’s secrets

ChatGPT is dumb. Seriously.

Thinking that it can spit out great content without clear instructions is like expecting a garden to bloom without watering it.

But here’s the catch: if you want to be clear about what you want, you need to know what you want.

Re-read this sentence again.

It’s not just about knowing that you want to create a great blog post. You need to know what actually makes a blog post stand out. And how do you know that? By practicing your craft.

You see, so many people think that AI is a shortcut to becoming lazy, rich, and famous.

But it’s not.

It will not replace learning and experience. Sure, it’ll make things faster, but you’ll always be limited by your own boundaries.

Here’s how to think about it:

You need to treat your AI like a headless assistant. Yes, it types faster than your hot little fingers. But it doesn’t have your intuition, your years of experience, or your in-depth knowledge.

So, work on your expertise and become better at your craft.

How would Warren Buffet use ChatGPT?

By now, you get it. It’s all about prompts.

Engineering prompts can be a huge time saver. But they can also be huge time suckers.

I’ve been there, pouring hours into perfecting a prompt, only to find that I would have been faster by simply writing the damn thing myself. The sinking feeling in the pit of my stomach served as a stark reminder of time wasted.

Let me put this into perspective.

Just like you’d evaluate the ROI before automating a task, do the same for prompt engineering. Weigh the time spent against the quality of content and the potential benefits.

Ask yourself if the content piece you want to write will:

  • Make your content workflow faster?
  • Help you spend less time in front of a computer?
  • Make you engage with your audience on a deeper level?

As an example, I’ve invested over 15 hours in prompts to write emails to launch my courses.

Given the fact that with my current templates, it only takes me 15 minutes to write an email… I’ll need 60 emails to break even on the time spent. Only after email number 61 would I start getting a positive ROI.

In my case, that’s an ROI worth pursuing because I’m launching a lot of courses and writing a lot. But it might not be the right thing for you.

Invest time in prompt engineering only when there’s a clear return. Otherwise, you’re just wasting your freakin’ time.

Always think in terms of returns.

Final words

AI ain’t a magic bullet if you have no clue what you’re doing.

Skyrocketing your productivity with AI is only possible when you put in the reps to design great prompts that reflect your experience and your voice.

Experience. Learn. Share.

FREE Email Course Reveals


Your data is safe. You can unsub anytime.