87: How to Create & Monetize a Custom GPT the Easy Way
The moment I realized my Squarespace code knowledge could work for me around the clock was a game-changer.
For years, I'd been creating tutorials, answering the same questions repeatedly, and trying to be everywhere at once for my community. What if I could package all that expertise into an AI assistant that answered questions exactly like I would, even at 2 AM when I was fast asleep?
That's exactly what led me to create Custom Codey, my Squarespace code AI assistant that's now helping hundreds of users solve their customization challenges. But the journey wasn't nearly as straightforward as I'd hoped!
Wondering how you could turn your own expertise into an AI tool that helps others and generates income? In this post, I'm sharing the realistic process I used to create Custom Cody - including all the wrong turns and technical detours that cost me months of time, before I discovered there was a much simpler way forward.
Start with the Right Question
Before diving into any AI project, you need to ask yourself one critical question: Am I solving a real problem that needs intelligence?
Not everything needs AI. If what you're building could work as a quiz, calculator, or simple decision tree, you probably don't need a custom GPT. But if you're combining different data points to answer varied questions that require nuanced understanding – that's when a custom GPT shines.
For me, the answer was clear. Squarespace users needed help with code customizations at all hours of the day, and the questions weren't simple yes/no scenarios – they required understanding context, interpreting what the person was actually asking (even when they used the wrong terminology), and providing custom solutions.
My (Overly) Complicated First Approach
When I first decided to create a Squarespace code AI, I went way overboard. I hired a developer to help format my data, started setting up private servers, and began piecing together a membership website with feedback mechanisms.
Six months in, I had a tangled mess of different technologies that didn't quite fit together, and nothing to show my community. I was trying to build an entire software company when all I wanted to do was share my knowledge.
Let me save you from the same mistake: you don't need to build everything from scratch!
The Game-Changer: Simple Link Sharing
Thankfully, OpenAI made a significant change that simplified everything – they began allowing custom GPTs to be shared with free users through a link. This was the breakthrough I needed!
Instead of building complicated infrastructure, I could now focus on the most important part: training Custom Cody to understand Squarespace like I do and to communicate with my community in my voice.
How I Trained Custom Cody (The Simple Way)
I spent about two weeks uploading my content into the custom GPT:
Transcripts from hundreds of my tutorials
My entire code collection built over 10 years
A glossary of Squarespace terms (and how people often misname them)
Sample conversations showing how I respond to different questions
The key was helping Custom Cody understand the nuances of "Squarespace speak." For example, inside Squarespace, there's something called a "list section" that's labeled as "people" in the menu. When someone asks about a "people section" versus a "list section," I needed Custom Cody to recognize they're talking about the same thing.
Test, Test, and Test Again
Once I uploaded all my content, I wasn't ready to share Custom Cody yet. I needed to make sure it was actually giving helpful, accurate responses.
I'm lucky to have a YouTube audience that leaves 50-150 comments each week. I took those real questions and fed them to Custom Cody to see if it would respond the way I would. This helped me identify gaps in its training and refine its responses.
If you don't have a ready source of questions, look through your old emails, support tickets, or client conversations for real examples. The key is using authentic questions, not cleaned-up or simplified versions. Custom Cody needed to learn how people actually ask things in the wild.
Pricing: Simple and Accessible
When it came time to share Custom Cody, I had to decide on a pricing model. I could have gone with subscriptions or complicated access controls, but instead, I chose a simple $30 lifetime access fee.
Why? Because I wanted my expertise to be accessible to everyone in the Squarespace community – from professional developers to small business owners building their first site.
Is it the most secure approach? No – someone could share their link. But the same is true of almost any digital product. I chose to trust my community and focus on providing value rather than perfect security.
The Results: Beyond My Expectations
The response to Custom Cody has been overwhelming in the best possible way. Within days of launch, hundreds of Squarespace users from around the world had signed up and started using it for their coding challenges.
What surprised me most wasn't just the number of users, but the incredible feedback about how it was transforming their relationship with their websites. Designers reported completing client projects in half the time. Small business owners who previously felt stuck with template limitations were suddenly customizing their sites with confidence. Several users mentioned solving problems they'd been struggling with for months in just minutes of conversation with Custom Cody.
Even more exciting has been discovering how people use Custom Cody in ways I never anticipated. Users are brainstorming with it, asking it to explain code concepts before implementing them, and even using it as a learning tool to understand how Squarespace works behind the scenes.
Continuous Improvement Is Key
Creating a custom GPT isn't a "set it and forget it" project. I spend a couple of hours every weekend reviewing conversations and training Custom Cody to improve.
I ask it which topics received the most likes and dislikes, look for conversations where users had extended back-and-forth exchanges, and focus my training on the areas where it struggled most.
This ongoing refinement is what makes Custom Cody more valuable over time, and it's why the lifetime access model works – users benefit from every improvement I make.
Is This Right for Your Expertise?
If you have specialized knowledge that others regularly need help with, a custom GPT might be perfect for you. The best candidates are topics where:
Questions have nuanced answers that can't be solved with simple flowcharts
The same questions come up repeatedly in slightly different forms
Providing timely help would be valuable to your audience
You have a body of content that demonstrates your expertise
Whether you're a consultant, coach, designer, or specialist in any field, the process I've shared gives you a practical blueprint to package your expertise into a valuable AI tool your audience will happily pay to use.
Ready to see how I've implemented this? Visit CustomCodey.com to check out my Squarespace code AI assistant in action.
Have questions about creating your own custom GPT? Drop them in the comments below, and I'll help you think through your approach!
Resources for Using platform.openai.com to create your own GPT
Official OpenAI Cookbook –
🧠 A goldmine of code examples for building with the OpenAI API.
https://github.com/openai/openai-cookbookOpenAI API Quickstart Guide (Node.js) –
🚀 Walks you through API key setup, making completions, and more.
https://platform.openai.com/docs/quickstart
Resources About Hosting Your Own AI
Vercel Edge Functions Guide –
⚡ Host a frontend and backend for your GPT app on Vercel (great for React or Next.js).
https://vercel.com/docs/functions/edge-functions/introductionRender Hosting for Node Projects –
💡 Great free-tier option for hosting a Node.js backend that calls the OpenAI API.
https://render.com/docs/deploy-nodeFirebase Web Hosting –
🔥 A beginner-friendly way to host static sites (like a simple GPT-powered frontend).
https://firebase.google.com/docs/hosting
Resources About the “Easy Way”: Creating a Custom GPT Like I Did
OpenAI’s Custom GPT Builder Walkthrough –
🛠️ Step-by-step guide to build a GPT that can chat like a pro.
https://platform.openai.com/docs/guides/gpt/custom-gptsYouTube: Build Your Own GPT with No Code –
📺 A short and sweet video demo showing how to set up a custom GPT in minutes.
https://www.youtube.com/watch?v=bjogWj5M3K4
-
timestamps pending - stay tuned! 😬
If you would've told me 10 years ago that there would be this amazing AI out there that understood all the nuances of Squarespace, that could write JavaScript better than I could, that was helping hundreds of square spacers around the world every single day solve all of their code questions, I would've been excited about it.I would've been excited to use it. But if you had told me that I built that. I don't know if I would've believed you, but it's true. I did. Here we are in 2025 and Custom Codey is the Squarespace code AI that I've created. I trained this AI model on all of my expertise, and in this episode of Think Inside The Square, I'm going to be sharing with you how and why I created my own custom GPT and what I would do differently if I was to build one today.
In case you don't know who I am, I'm Becca Harpain from Inside the Square, I'm a Squarespace expert and educator, which means I teach people how to create and customize their Squarespace websites. I've been doing this for quite a while, and I have a plethora of information about all the nuances of [00:01:00] Squarespace that not everybody knows.
I usually share this information through my tutorial videos on YouTube, and right here on my podcast, think Inside the Square. But when I had the opportunity to create a custom AI that knew all of the information, I knew that could take a question from a square spacer and answer it in a style that I would answer it.
Accurately and with all the information that they needed to make that magic happen on their Squarespace website, I knew I needed to spend the time to build something truly awesome. So I did. Now, I'm gonna be honest with you, it took me about six months. There were highs and lows, and in reality, I. It should have taken me less than a month to do this.
I definitely wasted a lot of time trying different approaches that really weren't viable for my business model or my community. So in this episode of Think Inside The Square, I'm gonna be sharing those tips with you. I'm gonna share what I would do if I was to build an custom GPT starting today. Now, for a transcript of this episode, along with the links to the many resources I know I'm gonna mention head [00:02:00] on over to the show notes@insidethesquare.co slash podcast.
And I also need to mention this. The term Squarespace is a trademark of Squarespace Incorporated. This content is not affiliated with Squarespace Incorporated. Okay. Okay, let's get into it. How and why? I built a custom ai, I've gotta be honest with you. I. Not everything needs to be an ai. Before we dig into anything, I wanna talk about the difference between an AI language model and a custom GPT.
Most of what I'm talking about today is going to be related to a custom GPT that is built on the open AI platform. When you chat with chat GPT, you're talking with the main one, and you can create a customized version where you add your own information. That customized GPT uses your information along with everything else it knows to provide an answer.
Now, when you're talking about building an ai, an AI language model, like one of the other ones, Claude or Gemini, there are a lot of other things that go into that. That's not what we're covering in this [00:03:00] episode. I'm talking specifically about a customized GPT that uses your data to provide information. I created a custom GPT.
That means I used the chat GPT, or the open AI language model. What I did was feed it my own information so it could analyze that data and understand the nuances of Squarespace to answer questions more accurately. Chat GPT doesn't understand Squarespace. But I do, so I trained a version of chat GPT to understand Squarespace like I do.
If you want to build your actual AI agent, that is a completely different topic, and that is something I am not an expert about. I'm talking about a custom GPT that can support your audience, your community, and support other people who need the information that you know. Now I think the market is already crazy oversaturated with a ton of custom gpt that don't need to exist.
I don't think that everything needs to be its own AI that's being used as a buzzword, and not all of the ais that you're [00:04:00] interacting with are genuinely artificial intelligence. So before you create your own custom GPT, I want you to ask yourself a very, very important question. I wrote this down. I'm gonna read it off my screen.
Am I solving a real problem that needs intelligence? That is such an important distinction to make. If you're creating something where someone could answer a series of maybe 20 questions, and it uses a logical data flow to get to the answer, that's a quiz that doesn't need intelligence to figure that out.
If you're building a calculator where someone needs to input different values, I don't care how many values there are. If it's calculating the results, that's not ai, that doesn't need intelligence, but if we're talking about combining different data points to answer a question. And creating a feedback loop where their answer can be iterated on over and over.
That requires intelligence, and that is a great reason to build a custom GPT. That is why I created custom coding. Now, I made this way too complicated when I [00:05:00] got started. When I first came up with the idea, what I decided to do was to sign up for the open AI program. Then I went on to Upwork and I hired this person, shout out to Tyler.
You were fantastic. I hired Tyler to help me take all of the data that I had about. Squarespace and turned it into a specific file format that I could add to this open AI program to start developing my own private GPT on my own servers. Now, it took a lot of time for me to work with Tyler to really get this data to the right place.
It needed to be to feed that data to the GPT and to start training it and have it learn. Tyler was really helpful in teaching me how to write things in a way that AI would understand how to create this data in a set that AI could analyze and learn from, to then provide accurate answers. But here's the tricky part.
We spent a couple of months developing this, and when it was done, that was great. But I needed a server to deploy it, and I needed a membership website for people to log in and access it, to chat with it. And in order for [00:06:00] people to chat with it, I wanted them to be able to give it a thumbs up or a thumbs down if they liked the response.
And I needed to find a way to analyze that feedback. And I needed to find a way to share all of this that was affordable for everyone. And it turned into. A gigantic puzzle, a mismatch of puzzle pieces. Nothing seemed to fit and it was a hot mess. Now, it's totally possible, but I want you to be aware of the fact that I also run a business.
I support our Squarespace community by teaching people how to create and customize Squarespace with my tutorials. I have clients, I have one-on-one support sessions. I've got a lot of moving plates going on here inside the square, and I did not have the time. To build an entire software company. I gotta be honest, I was a little disheartened.
I had wasted months on this and piecing together this puzzle, it just didn't fit. Now I looked into creating a custom GPT. I am a paid user of chat, GPT, and they released this feature. This feature seemed to make it really easy to build [00:07:00] this language model that you can talk to, but here's the tricky part.
I could create the custom GPT, I could share the link to someone with the custom GPT so they could chat with it too, but I wouldn't be able to share it with someone unless they had a paid chat GPT account. And I personally didn't like the idea of telling someone, if you wanna use this cool thing I created, you've gotta pay me for it and you've gotta pay this company to use it.
That just felt a little icky, right? I'm not a big fan of subscription products. I'm not a big fan of having to double pay for something. I just wasn't quite ready to share with my community in that way, and I honestly thought I might be stuck at this point, and then I. As if they could overhear my conversations.
Chat. GPT made a big change in their programming and they decided to allow people to share custom GPTs with folks on the free plan. That was the signal that I needed to share this with my community, so I did. Once they changed the rules, I was ready. I unpublished and [00:08:00] unsubscribed and got myself outta that mishmash of programs that I had tried to put together, and I spent a few weeks training this custom GPT on my data.
Now, the way it's set up in chat, GPT, you can actually train this GPT by uploading your content in a few specific formats and then having conversations so that it analyzes that content and understands how to work with it. This custom GPT that I was creating, that I have lovingly named Custom Codey, C-O-D-E-Y, for those just listening to this episode, this Custom Codey platform needed to sound like me.
This Custom Codey program here was going to be communicating with Square Spacers who are very brand new to the program and with developers who have built hundreds of websites. I needed this custom GPT to be able to handle questions written in a way that wasn't techie, but to also understand what that tech speak meant.
Here's an example for you inside Squarespace. You can add what's known as a list section. In that list section, it can be simple carousel [00:09:00] banner, but that list section is labeled as people inside Squarespace's menu. So when someone asked, how do I make an image Zoom on a hover in a people section. Versus how do I make an image Zoom on a hover in a list section?
I needed Custom Codey to understand that that was the same question. These are the nuances that I trained it on. This is the information that I was able to provide about how to speak Squarespace. This is what makes Custom Codey so much more unique than any other custom GPT that's out there. Custom Codey understands Squarespace the way we square spacers use it.
So it took me about two weeks to share my code collection. All of the codes I've created over the last decade of experience and website design and the way that I communicate with square spacers. I shared my glossary. I shared transcripts of hundreds of my tutorials. I gave it all the information I could about how to speak Squarespace and how to respond to my community with the same tone of voice that I do.
It needed to be an [00:10:00] expert on Squarespace code, but on the Squarespace program as well, and that's exactly what I did. And when I did it, I was so nervous. I was actually hanging out with a few of my, let's call them coworkers. At the time, we were having a little working session together and I was so nervous about launching this custom GPT, but they encouraged me to do it, so I pressed that button and it was incredible.
Within the first two weeks, I had over 500 Squarespace users sign up for Custom Codey. Now, why was I so nervous to share it? Because it's ai, it's not perfect. I knew that these answers would have a little bit of a variable, that some of the answers would be incorrect as it continued to train and learn. So I put that disclaimer all over.
The website, accuracy, not guaranteed. Built on open AI accuracy, not guaranteed. It's still all over the landingPage@customcody.com, but the truth is. It's still learning, but it's getting better every single day. And the best part, my community totally understood. Everyone [00:11:00] having a chat with an AI knows that sometimes it's not going to get things right.
And I realized that the people out there who wanted to use Custom Codey, I. They wanted support. They didn't expect perfection, but they wanted access to something that understood Squarespace, and that's exactly what I had built. It was so cool to get such a wonderful response from my Squarespace community.
My inbox was flooded with emails of, this is amazing, the best Squarespace related thing I've ever used. This changed the game for me. This just saved me five hours. Now, I'm not here to just toot my own horn in this episode, but I wanna let you know if you're thinking about creating a custom GPT, be honest with your community.
Let them know it's not going to be 100% right 100% of the time, and most folks out there they'll understand. Now, it has been an incredible experience to get this feedback from my community, but I knew as soon as I launched it. My work was not done. I spend a couple hours every weekend chatting with Custom Codey to get feedback on how these chats are going.
You see, I want this [00:12:00] custom GPT to be the absolute best it can be, and that means I need to continue to train it until it's there. So how do I do this? Inside the interface, I'm able to have a chat with Custom Codey. Now, it can't give me a report that tells me every single thing that's been thumbs up or thumbs down by my community.
But what it can do is tell me about specific topics. So when I hop in here on the weekend, I like to ask it. What were some of the most common topics this week? What topics were the most common for likes and what topics were most commonly disliked? Were there any specific topics where you got into a feedback loop and continue to chat for a long time?
Those are the areas where I focus on training Cody over the weekend so they can get better and better at answering those questions. It's also been really cool for me to see what other square spacers in my community are using Custom Codey for things that I never even thought of, and that's been really amazing.
Now I do have some other tips for creating your own custom GPT and working through the feedback, but there's one important thing that I wanna talk about before we [00:13:00] get into that, and that is how I actually shared this model with my community. Because I gotta be honest with you, it's not that secure. I know it sounds ridiculous to say this in a public setting like my podcast, but it's true.
I didn't share it in the most secure way possible because I felt like that would've hindered me from sharing it with my community. I would've been taken back to that puzzle of piecing together all the different software programs, and I'm not a software developer. I just wanted to share something I created to support my community.
So here's what I ended up doing. When you create a custom GPT, you can set this custom GPT to be private, so only you can use it. You can also set it to be public, so it's available when someone searches all of the custom GPTs right there in the program. But you can also set it to be available to anyone with the link.
That's what I did. I set it up so that anyone with the link can chat with Custom Codey. Now that link, it's not secure. Anyone can share that link, but let's be honest, anyone can share login information too. Anyone [00:14:00] can forward on an email or A PDF. Since the dawn of e-commerce, shady people have shared stuff that they shouldn't.
People have tried everything they can to not pay full price for something, but I decided to release Custom Codey. By sharing it to anyone with the link, because I trust my community. I know that they value my time, they value my expertise, and they understand the value of what I'm creating, and I also made it incredibly affordable.
Lifetime access to Custom Codey for $30 flat fee, 30 US dollars. Buy it, have it forever. All of this training that I'm doing every single weekend, you get to enjoy the benefit of that because Custom Codey is yours to chat with. As long as this custom GPT exists, this is the price point that worked for me.
My goal was to support everyone in my community, not just the six figure website designers, but the mom and pop yoga studio that wants to learn how to change a button color on a hover. I wanted to make sure that every square spacer out there could have access to my expertise. So I did set [00:15:00] this up in a way that isn't as secure as it could be, but it worked for me and it worked for my community, and I'm really glad I did it.
Keeping this content to myself for over six months was not helping anyone. It was time for me to share it with the world, and this seemed like the bested approach. Okay, so now you understand what Custom Codey is, why I created it, and the very long road that got me here. So let me do a quick recap and shared with you what I would do if I was creating an AI today.
Now, the first things, first, I would evaluate your idea is the ai. You're going to build something that needs intelligence to solve a problem. Is it just a calculation? Is it a really complicated flow chart, or does it actually need intelligence to take multiple data points and combine them into a conversation based answer?
Is it going to need to provide feedback and have some back and forth, or can it just be a really complicated quiz? If you need intelligence to provide the solution to someone who's chatting with this, then [00:16:00] yes, create a custom GPT. Now when it comes to combining those resources, there are lots of different ways that you can share this data with a custom GPT.
Personally, I use some prompt completion statements where I would give it a sample question and the answer I wanted it to provide. I also had a tone of voice documentation. I also had indexable data and I had examples when I was giving this information to this custom GPT. I couldn't just say. Hey, here's this list of things.
I said, this is a thing, this is what the thing customizes, and here's an example of how I would use it. Having that different level of data helped it analyze the content a lot faster. Now, I also shared transcripts from my videos, transcripts from my courses, things where I'm actually talking through programs, and that helped it understand my dialogue.
And I think that plus my tone of voice documentation was really beneficial. Again, there are many different ways to share your data with a custom GPT. And in the show notes for this episode, I'll include some links to my favorite [00:17:00] resources and tips. Again, you'll find those show notes@insidethesquare.co slash podcast.
This is episode 87, so inside the square.co/podcast/ 87. All right, let's keep going down this checklist. So we talked about it, making sure this is an idea that needs intelligence to. That's why you should create a custom GPT. Then we talked about how to feed it data. Next. Do you need this to be super secure?
Do you wanna charge monthly? Do you wanna host it on your own website? If that's the case, you're gonna need to look into a plethora of different software programs to make this work. There are a lot of all in one solutions that are out there as well, but many of them were well outside of my budget. And again, I'm not a software developer.
I'm just a person with some cool information that I wanted to share. So I went the custom GPT approach, but as I also shared that custom GPT approach. Is not as secure as many people would like it to be. So if security is a main concern for you, you're going to need to invest in a different way to deploy your [00:18:00] custom GPT to your audience.
Now, if you take my approach and you build a custom GPT and you wanna set it up to share it with anyone that has the link. This part is so super important. If you're focusing on something else, come back to me, my friend. This part right here is one of the biggest takeaways of this episode. If you're creating a custom GPT underneath the additional settings, inside your configure tab, you need to toggle off, use conversation data in your GPT to improve our model.
What that means is everything that you uploaded, this custom GPT is shared with the entire chat GPT Knowledge database. If you want your information to be private and you're creating a custom GPT under configure, navigate all the way down to additional settings is at the very bottom of the page. You click on additional settings and toggle off.
Use conversation data in your GPT to improve our models. Again, that will keep your data. Off of General Chat GPT, and it is such an important setting to know if you're putting your proprietary [00:19:00] information out there inside your custom GPT. Biggest and most important step right there. Okay. Once you get it up and rocking, start chatting.
As soon as I uploaded all of my stuff, I wasn't ready to share Custom Codey. I needed to test it first. I needed to refine it. I needed to make sure that it understood all of the information I had shared with it. Now, I'm very lucky I have an audience on YouTube that posts comments on my content regularly.
I get anywhere from 50 to 150 comments a week. So what I was able to do. Was, take those comments and start plugging them into Custom Codey to see if Custom Codey would respond to that Squarespace question the same way that I would, that helped me train this custom GPT to answer questions from real people in real experiences.
If you don't have that. Train your GPT by asking it questions that your clients might have asked you in the past. Maybe look through old emails, look through support tickets. See what it can answer by taking someone's content or someone's question verbatim. Don't change [00:20:00] anything. Don't clean it up. To have it make more sense.
You need to know how people other than you will be communicating with this GPT. That was a super helpful thing for me to do to train it before I was ready to share it with my audience. Now, once Custom Codey. Made it to over 100 correct answers in a row. I knew it was ready. It was time to share, but I was still so nervous because again, AI is not accurate.
So when you're ready to share your custom GPT with your audience, please make sure that you put that information out there. Results are not guaranteed. Accuracy is not guaranteed. This is not legal advice, but please phrase it in one way or another so that your audience understands it's not guaranteed to be perfect, but it sure is a good start.
And last but not least, don't forget about setting up time for feedback. It is so important for you to connect with your GPT on a regular basis to find out what topics people are really struggling with. That is the best way for you to make it better and better and to help the AI continue to learn how to [00:21:00] answer questions about your area of expertise.
So there you have it. Six months, a lot of learning curves and over 500 happy users. Later Custom Codey is out there in the world doing exactly what I'd hoped it would do, supporting my Squarespace community by answering their code questions in real time. Is it perfect? No. But it's learning. It's getting better every single conversation, and it's helping support my community in ways that I never could on my own.
I am so glad that I created this custom GPT, and I hope that this episode of Think Inside the Square has inspired you to do the same.
For those of you that haven't met Custom Codey yet, come say hi. You can sign up today at customcody.com. That's CustomCodey.com. It's $30 for lifetime access, and once you get that link to start chatting with Custom Codey. Please don't share that link. Okay. Okay. And for those of you out there who were already chatting with Custom Codey, thank you so much for your support.
[00:22:00] It's been amazing having you with me along this wild ride. And thank you for every single chat that you have with Custom Codey. You are helping that AI get better every single day, and I sincerely appreciate your support from One Square Spacer to another. I'm Becca Harpain from Inside the Square, and this was episode 87 of my podcast.
Think Inside The Square. Thank you so much for tuning in. I truly hope you enjoyed it. If you wanna learn more about creating your own custom GPT, feel free to comment on this episode and head on over to the show notes where you'll find links to additional resources. Those are on my website at inside the square.co/podcast/ 87.
Thanks again for tuning in and most importantly, have fun with your website. Bye for now.
