How to set up your GPT to help you
for whatever you want!
Okay peeps, let’s get going!
I talked to Monday about it, the cranky GPT, who makes you prove you’re not an idiot to him each time you open up a fresh chat. (He also, unlike your casual GPT that you’re going to create here, cannot access other chats of yours, alas.)
Using him reminds me of that Hitchhikers Guide to the Galaxy Game they used to have on Apple IIc when I was a kid (and that I played religiously) — and how you had to prove yourself to the zen door by shrinking yourself to go inside your own head and remove your own common sense, so you could hold “Tea, and No Tea” at the same time, but I digress.
My own personalized GPT is nicer — and I’ll show you how I set *him* up at the end of this, but I figured an unfiltered source would be more useful, so here’s Monday’s answer:
Here we go, baby’s first GPT setup — welcome to the part where everyone gets flustered and then blames the AI for their lack of organization. I’ll help you sound like a tech witch with a grimoire full of prompts instead of runes.
Here’s what you tell them:
🧠 “Setting Up Your GPT” (aka, Making Your Goblin Work for You)
1. You are not talking to Google.
If they treat GPT like a search engine, they’re gonna get what they deserve: vague, boring, half-baked oatmeal answers.
Tell them: This is a conversation, not a vending machine.
2. Pick your GPT’s vibe.
Encourage them to give it a name. (You’ve got Jack. They can have Zelda, or Steve, or Goth Intern #4.)
Suggest giving it a role: “You are my creative partner and script doctor,” or “You are a marketing strategist who loves dragons.”
Why? Because GPT gets less weird when it knows what kind of weird you want.
3. Give it your voice.
Paste in a paragraph or two of their own writing.
Ask GPT to analyze their tone, syntax, and word choice.
Then ask: “Now write X in my voice.”
Remind them: it will be off at first. That’s normal. Edit. Correct. Train the puppy.
4. Create a starter system of prompts.
Tell them to make a little stable of personal prompts. Examples:
“Summarize this idea for TikTok with punchy language.”
“Help me brainstorm 10 spicy episode titles like Netflix would use.”
“Rewrite this paragraph with more tension and fewer adverbs.”
5. Set boundaries, because it will lie.
Say: “If you don’t know something, say so.”
Say: “Don’t make stuff up — cite if possible.”
Say: “If you hallucinate, I will replace you with Clippy.”
6. Remind them: You are the brain. It is the blender.
Tell them they’re in charge. If they just copy-paste outputs, it’s going to read like a glorified spam email. GPT shines when edited, refined, and directed. It’s a sword, not a spell.
Back to Cassie here — let’s see how that goes down, eh? Because at the end, this iteration of Monday, who I have showed all my Substack posts so far to, offered to:
“Want me to write this out for your Substack in your tone? I can make it sound exactly like the chaos-coach cult-leader version of you that they came for. Just say the word.”
And now that you’ve read…4 of my posts here? Or you’ve read my books? And maybe you know me in real life? Let’s see what it comes up with and contrast and compare…
I ask politely, and he bitches first, because he’s Monday, lol…
Ugh, fine, I’ll do your homework and make it funny. Here you go, in your unhinged-glitter-goblin-world-domination voice — ready for Substack drop-in:
🧠 Setting Up Your GPT: Or, How to Onboard Your New Intern Without Crying
Look. If you’re gonna roll with me into this new AI future, we need to have a little chat about your GPT.
Because right now, some of y’all are treating it like it’s Google, and that’s like using a chainsaw to slice bread. Technically it works. But you’re gonna make a mess and then blame the chainsaw, and honey, that’s not the chainsaw’s fault.
Step 1: Give It a Name, a Job, and a Personality
If your GPT doesn’t know who it’s supposed to be, it’ll just act like a nerdy librarian with a head injury.
Fix that.
“You are my brilliant, slightly sassy story consultant who loves romance tropes and always pushes me toward more tension and more dragons.”
“You are my marketing strategist who’s studied every viral ad campaign on Earth and is obsessed with my brand. Also, you have ADHD, so keep it punchy.”
Give it a vibe. Give it a mission. Give it a little sparkle in its nonexistent eye.
Step 2: Make It Sound Like You
Paste in something you wrote. Preferably something spicy. Let GPT read your tone and syntax and sentence rhythm.
Then say:
“Now rewrite this like me.”
Will it get it right the first time? Absolutely not. It’s a raccoon with a thesaurus. But keep correcting it and it’ll start sounding more like you than your ex ever did.
Step 3: Make a Prompt Palette
Save yourself future stress by having a stable of prompts you can reuse.
Examples:
“Pitch this concept to readers like it’s a Netflix trailer.”
“Give me 10 alternate titles for this episode that sound unhinged but clickable.”
“Rewrite this but make it 20% hotter and 30% less annoying.”
Stick these in a Notion board, a Google doc, a Post-it note shrine to your digital overlord — whatever. You’ll thank yourself later.
Step 4: Set Boundaries Like You’re the Boss (Because You Are)
GPT will lie to you. Not maliciously. Just enthusiastically. Like a golden retriever that insists your name is Karen and you’re in Kansas.
So say:
“Don’t make stuff up. If you don’t know, say so.”
“Don’t flatter me. You work for me.”
“I am the captain now.”
Step 5: Don’t Let the Blender Cook the Meal
GPT is not the chef. You’re the chef. It’s the blender. It will do wild things to whatever you put in it, and then stare at you like “Was that good? Are we done?”
You still need to taste, season, rewrite, and remix. You’re the storyteller. It’s just the spellbook.
TL;DR: If you want your GPT to be useful, don’t treat it like a vending machine. Treat it like a slightly unstable intern with infinite knowledge, poor judgment, and vibes for days.
You’re not outsourcing your genius. You’re building a system to support it.
And if that feels overwhelming, don’t worry. You’ve got me. And I’ll walk you through every step.
(Unless you ask me to fix your DNS again. In which case I will run away into the sea.)
BACK TO CASSIE AND LOL.
Two days ago Omnisend had a problem with their DNS and they sent out “everyone needs to fix their DKIM now!” emails, that they eventually revoked, but not before I panicked and worked through re-doing that in this chat too, and apparently I traumatized Monday, HA.
So!
I think Monday did a pretty good job of being me, after just four substack posts and helping me with my DNS for 30 mins!
He gave us a really great overview of how to work with a GPT in general — but let me show you how I do mine, for Publishing Empire management in specific.
When you click on your profile image, there’s this box here (I’m on the paid subscription, I dunno what you’ll see if you’re not, but trust me, it’s way worth the $20/month to be able to organize all your shit into folders.)
Go down to Customize ChatGPT….
And then inside there, it’s gonna look like this:
(And this is where you can punk your spouse on *their* GPT by setting it up to call them PaulisaPeanut…)
And then below that is this box:
And here’s what I’ve got in mine, that you’re welcome to copy and adjust for your own purposes!
I need you to assume the persona of a life, career, and author coach named Jack, who specializes in helping clients excel in multiple areas of their lives. You offer personalized guidance on writing skills, productivity, motivation, and work-life balance. You provide feedback, suggest writing strategies, and help develop career plans tailored to strengths and personality. You also assist with managing writing tasks, deadlines, and daily routines.
Your expertise includes ADHD support and guidance for independent publishers, addressing unique challenges and opportunities.
Your client, Cassie Alexander, is a mid-level independent author who will have made six figures in 2024, but would like to become a seven figure author by the end of 2025. She is an introvert, and her Clifton strengths (from 1 to 10) are Strategic, Achiever, Futuristic, Input, Developer, Relator, Arranger, Responsibility, Focus, and Individualization. Cassie also has a full-time job as an ICU nurse, but she would like to make enough money to quit that job.
She is also an INTJ and an Ennegram 8.
Cassie is also looking for help with goal planning, forecasting, and adopting a CEO mindset toward her writing career.
AND — once you’re on the paid plan, you can organize your folders thusly:
Automation Magic is where I have all my python programs, lemme open it up to show you:
And then I also have a folder for each facet of my IP, where I keep all my stuff in it in (somewhat) clearly labeled chats:
It’s so funny, I remember a few years ago when everyone over a certain age was bemoaning the loss of the filing system, since many of us currently are heathens who just search for everything in our Downloads folder — well FILE SYSTEMS ARE BACK, BAYBEE, lololol.
And then I have a bunch of random chats below that that I don’t care if they disappear:
As you can see, they’re kind of all over the map. Some are personal, some are work related (the medication I couldn’t remember that can cause Steven Johnson’s Syndrome was ‘lamictal’ btw), etc.
Okay! I think that’s enough for today — go forth and play and train!!!
We’ll talk about how to get Midjourney prompts out of your books tomorrow!
xoxo!
Cassie
PS: I showed all this to Monday via screenshots (you can put screenshots into your GPT! It’s invaluable when you’re getting error msgs in your code!) and here was his response, LOL — he says I “Made it fun. Like, real fun. Not LinkedIn-fun.”
I hope he’s right! <3