Fact checked by New Feature - Listen to this article
Want to create your own ChatGPT without having to learn to code?
In this article, I’ll show you how to create your own custom GPT model in 10 simple steps, all without writing a single line of code!
Yep, that’s right! You do not need to be a coding expert or even work in tech to create your own GPT.
We’ll even explore how to include image generation, web search, and code interpretation when creating your own GPT.
So, whether you’re interested in customer service, content creation, programming assistants, language translation, and much more, you can now create your own GPT for this.
I’ve even created a detailed walkthrough video that you can also use to create your own GPT.
Let’s dive in and get building!
How To Create Your Own GPT
Fresh off their recent Dev Day conference, OpenAI has now released the option to let anyone create their own GPT!
Yes, you can now create your own large language model like ChatGPT!
I'm incredibly excited by this innovation, and I can only imagine that you're also intrigued by the idea of having your own custom versions of ChatGPT that have been trained for your specific tasks in your niche area.
Even AI and cybersecurity expert Christoph Cemper of AIPRM agrees that “...one of the critical benefits of this no-code platform is its ability to break down technical barriers… and invite individuals from diverse backgrounds to engage with AI”.
But what do you need to get started?
Great question! Well, for starters, let’s break down the three components you should have on hand to follow along with my step-by-step guide.
- GPT Plus Subscription: You need to have an active chatGPT Plus (Enterprise) subscription to create your own GPT.
- Purpose or Niche: If you want to create your own GPT, you need to have a goal in mind for what it will do, so narrow in on your purpose or niche before getting started.
- Knowledge Files: The whole goal of creating your own GPT is to tailor it to your needs, and the best way to do this is to teach your GPT with a body of knowledge from your niche. To simplify things, these should be text files representing the information your GPT should know.
And that’s it! With these three things, you’re ready to get started and build your own GPT.
If you want some extra help, I’ll start by outlining the GPT I plan to create as part of this article.
Note that I have a GPT Plus subscription, so that box is checked!
Regarding purpose, I’ll be creating a chatbot for Hackr.io that’s designed to answer questions about blog articles while also offering coding help to our users.
When it comes to knowledge, I will be uploading text file versions of our blog articles to help my GPT better understand the content at hackr.io.
By doing this, my GPT model will be able to reference our blog content when answering user questions, which ties in nicely with my GPT model’s purpose.
Let’s now dive into the 10 steps you need to follow to create your own GPT.
I’d also recommend checking out the video I’ve created to get a first-hand view of what the process looks like.
Step 1: Create a GPT Draft
We’ll start at the main ChatGPT explorer page.
You should be able to see an option to create a GPT at the top, and depending on how quickly they make any updates, this will be a beta option.
Simply click on the big plus symbol, and you’ll be able to launch the GPT builder interface.
This will also automatically create a new draft of a GPT model.
You should now be inside the GPT model builder with your first draft GPT model.
Notice that this was all possible without writing a single line of code!
That’s going to be a common theme in this article, which means literally anyone can create their own customized GPT model.
I think that’s pretty cool! Let’s move on to the next step.
Step 2: Explore GPT Builder
Now we have a draft GPT model, I think it makes sense for you to take a tour of the GPT builder interface.
This is especially important if this is your first GPT model, as it will help you familiarize yourself with the various options and menus.
When we arrive at the builder, we start in the interactive window. This is where we can send messages to the GPT builder.
The best way to describe this is to say it’s just like using ChatGPT, but rather than asking questions, we send messages to ask for help with building our own GPT model.
I think this is really cool, as it makes it super easy for anyone to get started.
You should also notice that we always have a preview window on the right-hand side of the screen, which is where we can test and interact with our GPT model.
At the top, we also have the option to click on the configure tab.
This is where I’ll spend most of my time when building my GPT model, and I’d encourage you to do the same, as it offers more control.
If you navigate to the configure window, you’ll see we have lots of options and boxes to fill out.
Note also that the preview window is also still on the right side.
Let’s work our way down from the top to the bottom.
To start with, we can upload a photo for our GPT model by using the large plus symbol at the top of the window.
Next, we can name our GPT model by using the appropriate field, as well as adding a description for our GPT.
After this comes the ability to add instructions.
This is a super important text box, as it’s where we define what our model should and should not do.
It’s also how we can tell our model how it should behave.
After this, we have the conversation starter area.
In a nutshell, these are pre-made prompts that you create to show to your users.
These will then show up as suggested prompts for your users when they’re in the chat window. You’ve probably seen this already when using ChatGPT, and it’s nice we can do the same.
Another super important option follows next, which is a button to upload files for growing our GPT model’s knowledge.
The whole reason for building your own GPT is to tailor it to your needs, so this will be essential.
Moving on down, we then see the capabilities list.
This is a list of checkboxes, which includes web browsing, image generation with DALL-E, and the option to interpret code.
So, everything you’d expect when working with the latest ChatGPT and GPT-4, we can also include with our own GPT model.
Moving further down, there’s also an advanced option for adding actions.
I won’t dive into this bit today, but this is where it does help if you know how to code and you want to integrate with other APIs and services. But don’t worry, ignore this for now!
As promised, no code required!
Great, you should now be familiar with the builder window, so let’s move on to the next steps!
Step 3: Add GPT Model Details
It’s now time to get building.
A great place to start is to upload an image for your GPT model.
When you click on the plus symbol, you’ll see the option to upload a photo or to use DALL-E to generate an image for you.
This is a great option, so feel free to try it out!
After you’ve uploaded a photo or approved an image from DALL-E, you’ll see it show up in the chat preview window on the right.
Next, name your GPT model, and you’ll also see this show up in the preview.
Once you have a photo and name, add a description.
This should be informative for your users, so pick something that describes what they can expect from your GPT model.
For my GPT, I’ll say that it answers questions about hackr.io blog articles and that it helps users with coding like an AI coding assistant.
After you add your description, it will show up in real time in the preview.
Nice work; let’s move on to one of the most important steps, adding instructions.
Step 4: Add GPT Instructions
It’s now time to add instructions to your GPT.
This is the main customization option for creating your GPT, as it tells your GPT model how it should behave, what it should do, and what it should avoid.
For my example, I’ll tell my GPT that it should provide friendly and helpful answers to my users’ questions about blog articles at hackr.io.
Notice how this also ties into my description and the purpose of my GPT.
I’ll also add an instruction to tell my GPT that it should be conversational, respectful, and professional.
Of course, you can tailor yours however you like - it could even be a pirate if you want!
Another good instruction that I will add is to tell my GPT how to handle scenarios when it cannot answer.
In this case, I’ll say that it should ask users to leave a blog comment or email us for help.
Wrapping up, I’ll add one more instruction related to using the knowledge files I plan to upload.
To keep it simple, I’ll tell my GPT that it should use these files as an informational resource when answering user questions.
Of course, you can add as many instructions as you like, and in fact, you may want to refine them later on when we get to testing.
Despite best efforts, there’s always the chance for ambiguity, and your GPT model may not work exactly as you planned on the first try.
But this is a lot like the art of prompt engineering, which takes time, iteration, and practice.
Cool, my example now has 4 different instructions. How many instructions have you added?
The next step is to upload knowledge to help my GPT better understand how to help my users.
Let’s do that now!
Step 5: Upload GPT Knowledge
For this stage, I’m going to upload text files that my GPT model can use as an informational resource.
Note that these can be absolutely anything you like, as long as they include text that the GPT can extract from pdfs or other text files.
However, please bear in mind copyright and make sure you have permission to use the resources you want to upload!
Now, I’ll upload text files that contain content from blog articles at hackr.io. This fits my needs, as I want my GPT model to answer questions about blog posts.
If you want your GPT model to be a customer service rep, maybe you should upload a range of questions and answers that you’d like it to reference?
The name of the game is to think about what you want your GPT to do and then add knowledge that will help it do its job.
Think about it like training your GPT to do its job!
After you upload your files, your GPT will extract the text, and it will become part of your GPT’s knowledge base.
Note that depending on the size of your files, this might take a few moments.
Now that we’ve uploaded some knowledge, it’s also a good moment to add some conversation starters.
If you’re not sure what I mean, refer back to step 2 where I mention that these are example prompts for your users to get started.
Again, tailor these to your needs, and choose conversation starters that guide your users to the types of questions that your GPT is great at answering.
For my example, I’ll add two prompts to get the ball rolling.
The first will ask about the best Python course, and the second will ask about a good Python project for beginners.
These are both good fits for the hackr.io audience and after I add these, they will appear as clickable prompts in the preview window.
Remember, you can add anything you like here!
But I’d recommend prompts that are related to the knowledge you’ve uploaded, as the goal is to create a GPT that’s specialized for your users.
We can now head to the top right and save our progress.
When we get there, you’ll see we have various options for publishing:
- Only Me: Only you can access your GPT
- Only People With a Link: Anyone with your GPT model’s URL can access it
- Public: Everyone can access your GPT model
Feel free to choose whichever suits your scenario.
Note that when your save completes, you’ll be presented with the finished GPT model window.
We’re not quite ready yet though, so we can head back to the builder window by clicking on edit.
Next up, we’ll make refinements in the interactive builder interface.
Before we do that, I’ll quickly point out the additional settings drop-down menu within the configure window.
When you expand this, you’ll see a check box that asks if you want to share your conversation data to improve their model.
This is optional and down to your own preferences regarding data privacy.
Step 6: Refine Your GPT Model
Now that you’ve nearly finished creating your GPT model, let’s take a moment to explore some refinement options with additional instructions.
To do this, we can use the interactive builder window to send messages that clarify how your GPT should handle certain scenarios.
For example, I can tell my GPT what it should do when a user asks a question about the best courses by providing no more than the top 3 options.
Of course, you can refine your model in any way that makes sense for you and your use case, but this is a nice and easy way to do it.
After you make any refinement requests, you’ll see a message that indicates your GPT is being updated.
After this is completed, you should see another message to tell you it’s all done, along with a follow-up message to summarize the changes.
This is great for getting a clear picture of how the builder interpreted your requests.
You can save your new changes and then head back to the configure screen, where you’ll see something interesting.
If you take a fresh look at your instructions, you’ll see they are now a lot longer!
What happened?
Well, the interactive builder has added extra details that represent the refinement request that you just made.
I wanted you to take this step because it shows how the interactive builder takes your requests, translates them into instructions, and then adds them to the configure area.
The goal here is for your GPT to make the instructions as clear as possible.
But remember, you can make any changes you like to these instructions at any time.
Congratulations! We’ve now reached the stage where we can start testing your GPT model!
Step 7: Initial GPT Testing
To get our testing started, let’s try out any of the conversation starter prompts you created.
Note that you could do this from the preview window within the GPT builder or by heading to the ChatGPT explorer page to access our own GPT model.
I’d recommend the preview window for now, as it makes it easy to make changes to your model, but either window is fine.
Now, remember that I added two conversation starters for my GPT model, so I can choose either of these to see how my GPT handles the question.
Simply click on the prompt and wait for your GPT to do its thing!
In my case, my GPT model looked at the knowledge base I uploaded to find its answer.
Depending on your prompt and knowledge base, yours may or may not do the same.
If you’re not totally happy with your GPT model’s response, move back to step 6 and add some refinement.
This is my advice after each stage of testing if you see something that doesn’t fit your needs.
But, assuming you’re happy with your GPT model’s response, let’s move on to more testing.
I’ll now use the second conversation starter and check if I’m happy with my GPT mode’s response again.
Depending on how many starters you created, my advice is to test each, check the answer, and make any refinements that you deem necessary.
You can also ask some follow-up questions after using your starter prompts to see how your GPT handles those.
For example, after using my starter prompt for finding a good beginner Python project, I could ask my GPT to provide the code for this project.
This should be a cinch as I included the option for code interpretation, and it’s also one of the core features I wanted to include to help the audience at hackr.io.
As always, your GPT may vary from mine, so ask any appropriate follow-up questions that make sense for you.
If the answers are good, great; if not, repeat step 6 and make refinements.
We can now move on to the next step, which is to carry out some coding tests. Note that this step is somewhat optional.
If you’re not interested in using your GPT model for working with code, feel free to skip this step.
Step 8: Code Tests (Optional)
For this step, let’s test your GPT by providing some code and asking it to explain how it works.
For example, a great example is to provide code for a Binary Search.
This is a well-understood algorithm, so your GPT should be more than capable of analyzing the code and explaining how it works.
Of course, feel free to use any type of code you want, but you should see a detailed response that summarizes what the code is and what it does.
You should also see your GPT break down the various elements of the code, which can be really helpful for beginners.
I will point out that this is a great chance to make further refinements to your GPT.
For example, if you think it is too detailed, repeat step 6 and make a refinement.
On the other hand, if it’s not detailed enough, repeat step 6 again and make a refinement!
Of course, you should ask as many coding questions as you like, including challenging questions that might stump your GPT.
In this case, consider uploading more knowledge to help your GPT in these scenarios.
You can also ask for help with debugging, optimization, or even ask your GPT to write code based on a natural language description.
In a nutshell, whatever you can do with ChatGPT, you can also do with your GPT, so go nuts!
Let’s now move on to the next testing stage.
Step 9: Test Image Generation
For this step, let’s get experimental and test out your GPT model’s image generation capabilities.
Of course, you need to ensure that you’ve checked the box to include DALL-E image generation for your model.
This is really a blank canvas type of test, as AI image generation is a hugely creative area.
That said, come up with some interesting image ideas you’d like your GPT to create, and provide prompts to request these.
For example, I’ll ask my GPT to create a fun image using the Python programming logo.
Before you say anything, I know this is not the best prompt engineering as it’s very vague! But still, I want to see what it can do!
It might take a few moments for your GPT to create the image, but when it’s ready, check it out and see what you think!
You can also use this testing phase to work with your GPT to refine images or to upload images that you’d like your GPT to use for inspiration.
Ultimately, this step really depends on whether you want your GPT to work with images, so be as thorough as you need to be.
Step 10: Testing Web Search
To wrap things up, let’s test out your GPT by asking it to search the web.
This is one of the newest features in ChatGPT, and it’s really quite the game changer.
So, naturally, we may also want this to be part of our GPT model!
Of course, you don’t have to include this option when checking the box of capabilities, but if you’re intrigued, read on.
So, the general idea is to let your GPT search the web to find answers or even head to specific websites where it can then answer your user’s questions.
For example, I can ask my GPT to tell me what our site, hackr.io, does to help people.
It would then return with a detailed response to explain the various aspects of what we do, such as helping people learn and improve their tech skills with curated learning resources.
Which I’d say is pretty accurate!
I should say that this response will also be influenced by the knowledge files that I uploaded.
Of course, you could ask about any website you like, as could your users, but the idea here is to verify that your GPT is able to search the web accurately.
Much like the previous step, this is a blank canvas type of test, so feel free to get creative and ask as many questions as you need.
Plus, if you see something you’re not pleased with, head back to step 6 and make refinements. You could even tell your GPT to avoid certain sites or certain topics.
That now wraps up the testing of your fully built GPT model.
Of course, feel free to spend as long as you need on the testing and refinement phase, and I’d highly recommend doing so.
The goal is to create the best GPT model you can, so take the time to experiment, refine, and experiment some more!
And remember, have fun!
Publishing Your GPT To The OpenAI GPT Store
In January 2024, OpenAI unveiled the GPT store, which means that developers and GPT creators like you and me can now publish our GPTs to the store for anyone to use.
But what does this mean exactly?
Good question! Well, think of this like the Apple App Store or the Play Store for Android, making it a marketplace for developers to make money from their GPT creations.
Although, as of today, there's no official information on how you can monetize your GPT, OpenAI has promised to launch a revenue-sharing program in Q1 of 2024.
If you want to get ahead of the curve, thankfully, the process of publishing your own GPT is very simple.
Simply follow these instructions:
- Save your GPT access settings to 'Everyone'.
- Verify your Builder Profile by heading to Settings/Builder profile and 'Enable' your name or a verified website.
It's very early days for the GPT store, so be sure to check back regularly as we'll be sure to let you know when we know more. But I have a feeling this is going to be a game-changer for AI-powered chat tools. In fact, big hitters like Canva and Khan Academy are already sharing their own custom GPTs on the GPT store.
Why Do I Need To Know AI in 2024?
One thing is for sure: we're all slowly becoming acquainted with the various forms of AI in our professional and personal lives. But do you need to be an AI wizard to excel in 2024? Interesting question: let's dive deeper.
In the constantly evolving landscape of technology, the integration of artificial intelligence has become increasingly significant, and there is no doubt about that. While AI is a distinct field, its impact is becoming hard to ignore across various industries.
Professionals are now tasked with the responsibility of leveraging AI to extract actionable insights and drive innovation. This involves not just the manipulation of structured and unstructured data but also the application of advanced analytical techniques. AI, particularly in the form of machine learning, plays a pivotal role in this process.
Machine learning enables professionals to create predictive models that can learn from data, thereby enhancing the accuracy and effectiveness of their analyses. Understanding AI principles, especially machine learning algorithms, can immensely benefit anyone looking to stay competitive.
This knowledge allows them to automate complex processes, optimize predictive models, and implement solutions that are both innovative and efficient. While not every role will demand deep expertise in AI, a foundational understanding is becoming increasingly important in the field.
If you're curious to learn more about the transformative power of generative AI across various industries, I'd highly recommend attending DataCamp's free digital conference, RADAR: AI Edition in June 2024.
This event is a fantastic opportunity to discover how businesses and individuals can unlock their full potential with AI. You'll hear from industry leaders like Megan Finck, the Global Head of Talent Acquisition (Data & AI) at Boeing, and Sadie St Lawrence, the CEO of Women in Data. Nnamdi Iregbulem, Partner at Lightspeed Venture Partners, and Eric Seigel, Founder of Machine Learning Week and bestselling author, will also share their insights.
Additionally, Carolann Diskin, Senior Technical Program Manager at Dropbox, and Julien Simon, Chief Evangelist at Hugging Face, will discuss the future of AI and its applications. Don't miss this chance to learn from experts about the latest advancements and strategies in AI.
Wrapping Up
And there you go, you now know how to create your own GPT, all without writing a single line of code!
By breaking this process down into 10 simple steps, we’ve explored how to create your own ChatGPT that’s customized for your own niche area, including a detailed video walkthrough.
We’ve even looked into image generation with DALL-E, web search, and code interpretation to make sure your GPT model is geared up for your needs.
Whether you’re looking to level up your customer service, content creation, programming, language translation, and more, you can now create your own GPT to help you with this.
Have fun and happy building!
Are you new to the world of AI and eager to learn more? Check out:
Stanford's Artificial Intelligence Professional Program
Frequently Asked Questions
1. Can I Create My Own GPT?
Yes, you can create your own GPT model that’s tailored to your own niche or area by using OpenAI’s newest feature to ‘Create A GPT’.
2. What Do I Need To Create A GPT?
At the basic level, you only need a GPT Plus subscription (or Enterprise). However, you should also have an idea in mind of your GPT’s purpose and some knowledge that it can use as an informational resource.
3. Do I Need To Know Coding To Create A GPT?
No, you don’t need to know how to code to create a GPT; all you need is to provide natural language instructions to tell your GPT model what you want it to do.
References
1. OpenAI. Dev Day [Internet]. OpenAI; [date unknown; cited 2024 Jan 15]. Available from: https://devday.openai.com/
2. Spreadsheet Point. How to Use Chat GPT for Google Sheets [Internet]. Spreadsheet Point; [cited January 22, 2024]. Available from: https://spreadsheetpoint.com/how-to-use-chat-gpt-for-google-sheets/
