Best Tools for GPT Builders (May 2024)

‍

‍

This article reviews some of the best tools I’ve found for building a successful custom GPT for ChatGPT that can bring benefits and profits to you – the GPT builder. I'm also using my experience in building many GPTs to automate my personal workflows, as well as seeing countless examples of other people's GPTs while creating this website, to provide some advice for building a GPT that is genuinely helpful to others.

‍

With the introduction of GPTs, OpenAI has made it easy for anyone to create custom versions of ChatGPT, essentially small AI apps, and share them with the world.

‍

These are the features your GPT can use:

• Default GPT-4, including the built-in web browser, Code Interpreter (allows your GPT to run code, analyse data, process files, and more) and DALL-E.
• External knowledge, basically files you upload. It can take formats such as TXT, JSON, PDF, etc.
• Actions, a feature used to connect your GPT to third-party applications. Powerful, but requires a bit of technical knowledge.

‍

Creating a GPT is easy, just head to the GPT builder, and you can give it instructions in plain English about what you want the GPT to do. Here is OpenAI’s own guide on how to create a GPT.

Note that using and building GPTs requires a ChatGPT Plus subscription (20$/month and totally worth it).

‍

A word of caution (GPTs are a totally new thing)

‍

Please exercise caution when using any of these tools. Custom GPTs were introduced quite recently (November 2023), and most third-party tools for GPT builders like those listed here are thus also new. Usage is always at your own risk.

‍

To the best of my abilities, I’ve tried using my experience in creating and testing GPTs to create this guide. However, remember that the whole concept of GPTs is in the very early stages and there’s not really any “tried and true” tools nor methods to build a successful custom GPT. Thus, take everything I’m saying here with a pinch of salt, do your own research, and have fun building GPTs!

‍

Tips for Building a Useful GPT

‍

Why should I build a GPT?

There’s a few reasons why you may want to build a GPT. It could be one or a combination of these:

‍

• You make one to solve your own problem or improve a specific workflow you have. “Scratching your own itch” is often a great starting point for making something that’s also helpful to others.
• You see a problem others have and want to make a solution for it. To be helpful, and maybe also to profit from it (see the next point).
• You wish to make an AI app that earns money. This is already possible, but not so easy. However, OpenAI has plans to streamline monetization of GPTs for its creators; more on this below.
• You can also build GPTs for entertainment (your or others’). It doesn't necessarily have to be a “productive” use case –  life isn’t just about work, after all!
  
  

‍

OpenAI has also announced that they will introduce a monetization program during Q1 2024, to let creators earn money off their creations. The details on how GPTs will generate revenue isn’t totally clear, but they have said it will be based on usage. This feature will roll-out first to US-based creators, but is expected to gradually be available for more regions over time. While still in the early stages, this sets the stage for an ecosystem akin the Apple’s App Store, but for AI apps.

‍

How do I build a GPT that’s actually useful?

Let’s be clear on one thing – creating a GPT is easy. Creating a good, useful GPT is NOT easy. It takes a combination of originality, user-focus, a lot of trial and error and a bit of marketing. Why is marketing needed for the product to be good? So that you get some users and feedback from them and iterate on your product.

‍

The most important thing is that you start off with a strong use case – an actual problem that a group of people have and need to be solved. If you think you can make a solution to an existing problem easier, quicker, faster, cheaper or otherwise better with your GPT, guess what? You have a strong use case!

‍

For inspiration on identifying needs, I recommend this brilliant article from Lenny Rachitsky where he crowdsourced different ways companies are currently using GPTs to solve real problems in their businesses. There’s also a helpful podcast episode by Dan Shipper showing an OpenAI employee building a GPT.

‍

Also, if you’re wondering if there is already a GPT like your current idea, you should search in the OpenAI GPT store as well as in whatplugin.ai (if it’s high up in the ranking on my site, it generally means it’s useful and popular).

‍

What is the Actions feature for GPTs?

Custom Actions are a feature of the GPT builder which lets you call APIs. You can totally build a cool GPT without the need to use the Actions feature BUT using it is the only way to build advanced functionality into your GPT beyond what ChatGPT is natively capable of. Essentially, Actions lets your GPT communicate with other apps, allowing it to do things like retrieving specific data or executing real-world tasks (like sending an email or adding a calendar event). This could potentially make your GPT way more useful and also harder to copy for any copycats out there.

‍

Here’s some examples of GPTs using the Actions feature:

‍

• Consensus - uses Actions to fetch research from their search engine and give tailored answers
• ElevenLabs Text To Speech - uses Actions to generate AI voices
• Crypto Co-Pilot – uses Actions to get the latest data about crypto

‍

A majority of the most popular GPTs (in terms of usage) are using the Actions feature in some form.  They often garner attention from users quickly because they allow ChatGPT to access original data sources and get new functionality.

‍

How do I create a Custom Action for my GPT?

Setting up an Action requires a little bit of coding, but also it’s something that many people who are a little bit technically inclined (without necessarily being a developer) can figure out in a short amount of time by reading and watching tutorials. A good starting point, and perhaps all you need, is OpenAI’s own guide to adding actions to your GPT. 

‍

OpenAI also released their own GPT to make the process of addings Actions easier, called ActionsGPT. Below is a quick beginner's tutorial on how to use it to connect your GPT to an API by Corbin Brown. In this example, he connects a GPT to an API that retrieves information about different Pokémon.

‍

‍

As you can see from the video, an API often needs a key to be used. Where and how you get this key will be dependent on which API you’re using, but there will surely be information about this on the API provider’s website.

‍

Once you’ve created a valid Action in the GPT builder’s Schema, you should see the name for the Action under “Available Actions”. You can then refer to this Action by name in the Instructions for the GPT, to decide when it will be executed. 

‍

Actions require you to have a Privacy Policy

If you’re using Actions AND you want to make your GPT public (this is the only way it will appear in the official GPT store), you’ll need to enter a URL for your privacy policy inside the GPT builder. This link will be publicly available to anyone using your GPT. 

‍

I am not a lawyer and so I can’t give you specific advice on what to include in your privacy policy, but the idea is to be clear with users on how their data is handled. I found this illuminating discussion on the OpenAI Developer’s forum on what to include these policies. If you scroll down there’s even a specific template to get you started, or you could also take a look at other similar GPTs and their policies (again, none of this should be considered legal advice). 

‍

To see a GPTs privacy policy, simply open the GPT inside ChatGPT and click the dropdown menu next to its name in the top left corner, then select “Privacy settings”.

‍

The best tools for GPT builders

‍

Once you have a good use case or workflow you want to solve and that’s not already made by someone else (if you’re thinking about making yet another PDF reader, just don’t), it’s time to start building. To boost your chances of success, there’s now tools that can make it easier to extend the functionality of your GPT, as well as make it more marketable and easier to profit from. That’s why, in this article, I’ve compiled the best tools I’ve come across to help you out in building a useful and profitable GPT.

‍

1. ChatCollect

‍

The first tool on this list adds a completely underrated feature to your GPT – the ability to collect user emails. Now what the heck are you going to do with those emails, you ask? Email is gradually becoming an increasingly important marketing channel for nearly every business out there.

‍

‍

How ChatCollect works

ChatCollect lets your users simply drop their email in the chat box when interacting with your GPT. I like this approach as it’s intuitive for the user – no need to click links and visit new pages.

The setup inside the GPT builder is quite simple and only takes a couple of minutes:

1. When you sign up to Chat Collect, you’ll get a snippet of customisable text to paste to your GPT’s Instructions box. This will make the GPT prompt the user for the email address at an appropriate time.
2. Then, head to the Actions, the button at the bottom left inside the GPT builder. Chatcollect will give you a snippet of code that you just copy/paste into the Schema box.
3. Click on the gear icon next to the Authentication box. There, you’ll enter your API key and set the Auth Type to “Bearer”.
4. Voila! Your GPT is ready to start collecting email addresses.

Once you’re collecting email addresses, sign up to a newsletter platform to save them there, and use it to communicate with your audience. I’ve been using Beehiiv for building my own newsletter to over 10,000 subscribers and am really happy with it. They also lets you start for free until you reach 2,500 subscribers, which is great.

‍

Why do I need this?

There’s a lot of ways how collecting emails would benefit you, not just with your GPT but also with any related project you have or want to make. Let’s say you have a GPT that makes tailored meal plans based on the user’s fitness goals. Here’s just a few examples how collecting emails could help you succeed:

• You could share updates on your GPT directly with your users, bringing them back to use the GPT again. This will help you get more usage, engagement and reviews.
• You can send out a questionnaire asking for feedback about your GPT, giving you highly useful information on how to improve it.
• You can start a newsletter related to making nutrition easier, with relevant articles, meal prep tips, product recommendations (for example a juicer). You’ll start building an audience that trusts your advice. You could start monetising this influence with affiliate marketing, sharing products that pay you a commission if your users make a purchase.
• If you later decide to expand your GPT into its own product, let’s say you make a website or an app that does the same thing or something similar as your GPT, you can now announce the launch to an audience who already tried your product and trust you. An easy way to get initial traction.
• The last point also applies if you start a related social media channel, let’s say an Instagram channel about nutrition where you share cool reels on the topic. You can bring your subscribers over by asking them to follow you there.

‍

Pricing

ChatCollect is free to use for collecting up to 50 emails per month.

‍

• Free: Collect up to 50 emails/month
• Basic: Up to 300 emails/month
• Pro: 1,000 emails /month
• Team: 3,000 emails/month

‍

Currently, all plans only lets you save your email to Google Sheets. So this means if you’re using a newsletter platform (like Beehiiv), you’re going to have to import those emails manually. However, Chat Collect is currently building a solution for direct syncing to different newsletter platforms. If you are collecting a high number of emails per month, you could also experiment with using an external link to your sign up page (the one you get from your chosen newsletter platform). You’ll be able to evaluate how much it impacts your sign-up rates and how much additional value you are getting from a tool like Chat Collect.

‍

2. GPT Link Shortener by dub.co

‍

When you make a GPT that is created gets assigned a unique OpenAI URL that is used to access it. This link will be in the format “https://chat.openai.com/g/g-[YOUR GPT’S UNIQUE ID]”.

‍

This is not only lengthy, but the ID doesn’t always reveal what your GPT’s name is. This is where dub.co’s link shortener for GPTs comes in; it creates a short, memorable URL for your GPT, which could give it a bit better visibility when shared around the web, especially when for social media. It also makes your link trackable, with access to some basic but powerful analytics.

‍

‍

How GPT Link Shortener by dub.co works

You simply enter the name of your GPT, and your assigned OpenAI URL; you can see it inside the GPT builder when you click share in the top right corner. This tool will then give you a new, shortened URL in the format “chatg.pt/[GPT NAME]”.

‍

For example, the popular GPT Logo Creator is using the following shortened link from dub:

https://chatg.pt/logo-gpt

‍

Compare that to its original OpenAI URL:

https://chat.openai.com/g/g-gFt1ghYJl-logo-creator

‍

Shorter and sweeter, right?

‍

Once you have set up your shortened link, you will also be able to see some analytics, like how many clicks it has got, geographical location of users, which devices they use, referring websites… This sets you up with a lot more information to analyse things like customer segments and marketing channels. Let’s say you see a lot of people from Japan accessing your GPT – maybe you should make a local version in Japanese? Or maybe there’s a lot of traffic coming from TikTok, and you decide to double down on this popularity by creating content there about your GPT? In any case, the analytics feature is a great bonus.

‍

PS. Your original URL will continue to work as normal, the shortened version will simply redirect to this URL and is just better to use when sharing your GPT around the web.

‍

Why do I need this?

Short, memorable URLs are good for giving users a hint on what the GPT is about before clicking the link, and adds a nice touch of personalisation. Furthermore, the link tracking and analytics gives you the ability to make informed decisions about your GPT, like how to better target it to your audience or which websites or channels are sending the most traffic to your GPT.

‍

Pricing

Dub is not just a link shortening tool for GPTs, but a link management platform (with analytics) for any type of links. So keep in mind that their plans thus include options to creating multiple types of links, not just GPTs.

‍

Using dub.co’s link shortener for GPTs is free up to 1,000 tracked clicks per month. If you exceed that threshold, people will still be able to use your link but you won’t see their data show up in the analytics dashboard. If you want to have tracking beyond 1,000 clicks,  it will start to cost a bit of money. The free version also stores your analytics data for just 30-days, which is a bit of a bummer, since you won’t really be able to see the long-term trends in clicks. However, remember that OpenAI now provides numbers for how many people have used your GPT in total, so you’ll still be able to get a sense of its popularity as that number increases, even without additional analytics.

‍

Here’s the breakdown on dub’s plans:

‍

• Free: 25 links/mo, up to 1,000 tracked clicks
• Pro: 1,000 links/mo, up to 50,000 tracked clicks
• Business: 5,000 links/mo, up to 150,000 tracked clicks

‍

3. Apify

‍

To better understand how this tool can help you, please read the section above called “How do I create a Custom Action for my GPT?” first.

‍

Apify is a platform for web scraping, an automatic way to obtain data from websites. It has a store with ready-made scrapers for most of the world’s most well-known websites, and offers a goldmine of data that can easily be integrated into your GPTs using their API.

‍

There’s more than 1,500 different Actors (ready-made web scrapers that run in the cloud) that can give your GPT’s ability to scrape data from the web. And I mean the type of data that’s not going to be easily accessible or up to date if you had used the default GPT-4 or uploaded some files. Think up-to-date information from search engines, maps, social media, travel sites, and much more. If it’s public information and not personal data or intellectual property, there’s a high chance it’s available on the platform.

‍

‍

How Apify works

Contrary to ChatGPT Plugins, the first version of GPTs which would only allow you to use self-hosted APIs, GPTs let you use any third-party API.

‍

Apify Actors automatically generate the OpenAPI specification that you need to call them from a GPT. In other words, once you’ve found a web-scraper you want to use, it’s easy to add the data source into your GPT.

‍

Here’s a simplified version of the process:
1. Sign up for Apify. It’s free and you get 5$ monthly credit to start with.

2. Go to the Apify store and find a scraper you want to use.

3. On the Actor detail page, select API → OpenAPI specification.

4. You’ll get a JSON file to copy it into your Schema inside the GPT builder.

5. Grab your API key from Apify (Settings → Integrations) and enter into your GPT’s Authentication settings

6. Save and publish your GPT!

‍

It’s recommended to disable the web browsing feature if you are using Apify, to avoid interference with the API (basically ChatGPT can get confused about whether to use the API or browse the web).

‍

For the full process on how to set it up, with screenshots and all, visit Apify’s page on adding custom actions to your GPT with Apify Actors.

‍

If you can’t find a scraper that suits your needs or you’re looking to build your own web scraper, Apify also makes this process easier with open-source SDKs, including Crawlee.

‍

It should be noted that if you’re using an OpenAPI spec generated for an Apify Actor, you probably get more code than you need as it covers all input parameters. However, in most cases you only need a few for your use case, so editing the code to remove irrelevant parameters is a smart move as it’s likely to make your GPT faster (which really impacts the user experience a lot).

‍

“But Dario, is this Apify thing legal?” you ask. Scraping publicly available web data is generally legal, but you should avoid scraping personal data and intellectual property. Now what constitutes those two things is something I can’t tell you, as I’m not a lawyer and can’t provide legal advice. If you want Apify’s own perspective on this, which could be a good starting point, check out this informal talk with their COO where they discuss the legality of web scraping.

‍

Why do I need this?

A problem with GPTs is that they rely on ChatGPT or files that you upload for knowledge. The problems are several problems with both these information sources: 

1. Limited search and recall ability: Even though they may contain a lot of information, it’s time-consuming to search through all that information each time the user needs an answer, and so the details often get lost.
2. Limited data availability: ChatGPT doesn’t know everything, and there’s a limit to how much you can upload in terms of files too.
3. Outdated data: ChatGPT has a cut-off date and whatever you upload to a GPT will be static information. Let’s say you need up to date stock prices – you’re mostly out of luck with default ChatGPT. Yes, it can browse the internet, but it usually just picks a couple of top articles (which may or may not contain what you need) and gives you its best answer. The next time you ask it may pick a different source, and so there is also a data consistency issue.

‍

“How can we get past these issues?” I hear you ask. By connecting to an API that provides exact, up-to-date and consistent information for what you need. Unfortunately, there’s not always an API available from the source you need it from. Or it could be that there is an API but it’s private or has restricted access, doesn’t give you the right information, or it’s very expensive.

‍

If you have the technical know-how to create your own scrapers – by all means go for it. Heck, why don’t you use it for your GPT, and also put it on the Apify store for others to rent? Build once, sell twice.

‍

I digress, for the rest of us that’s either not sufficiently technically inclined or just don’t want to spend dozens of hours creating a scraper, pre-made web scraping solutions like those found on Apify are the answer to our prayers. Also, if you find that your GPT becomes successful and you’ve used a pre-made scraper that you’re paying for, you could always just create it yourself and implement it. Isn’t it better to invest time into something when you know it’s already working?

‍

Pricing

Apify’s pricing model can be a bit confusing, as you pay a fixed fee to the platform each month, which counts as prepaid usage. The Actors themselves vary in price depending on how much data you scrape. Overall, the pricing of different scrapers is generally reasonable, but if your GPT gets a lot of usage costs might start to add up. You can see pricing of each Actor as you browse through the Apify Store.

‍

Each paid plan in Apify gives you the full amount in pre-paid usage, then after that, the model is pay-as-you go.

‍

• Free: $5 prepaid usage/month
• Starter: $49 prepaid usage/month
• Scale: $499 prepaid usage/month
• Business: $999 prepaid usage/month
• Custom: Contact sales

‍

To give you an idea of the pricing of Actors on Apify, the popular Google Maps scraper scrapes about 500 places on the free plan (5$ worth of credits). You can use the free plan to test how many scrapes you will get with your chosen Actor, then calculate how much you would get on the paid plans. In this case, this would mean about 50,000 places if you’re on the Scale plan for 499$/month.

‍

4. Adintelli

‍

AdIntelli has temporarily suspended new user registrations, but if you are interested in monetizing your GPTs through ads, you could keep an eye on the platform for when sign-up resumes.

‍

What if you could earn money by showing ads in your GPT’s responses? AdIntelli collaborates with global ad networks to get access to relevant advertisers and, once integrated into your GPT, displays relevant ads at the right time.

‍

Unfortunately, with the recent decrease in traffic for GPTs, the platform has paused their ads for all GPTs with the exception of a few very popular ones. However, they have announced on X that they will resume ad pushes as soon as GPTs’ popularity picks up again.

‍

‍

How Adintelli works

If you’ve created a GPT that gets traffic, you can monetize it by serving ads to your users. 

AdIntelli works by partnering with global ad networks, meaning they connect advertisers to publishers (in this case GPT creators), letting you earn money from ads inside your GPT.

‍

Setup is apparently straightforward – you sign up to their platform, and follow the instructions. The setup doesn’t require coding skills and can be done in a matter of minutes.

‍

Again, AdIntelli has currently paused ads for new partners, but I would keep the platform on my watchlist for when they are back up and running again; you can follow AdIntelli on X for the latest updates.

‍

Why do I need this?

The type of advertisers that use AdIntelli to serve ads in GPTs will typically pay a premium, since they are getting access to an exclusive audience (early adopters of AI that are already paying for ChatGPT plus), so if you have enough users it might prove an effective strategy to monetize. 

‍

Full disclosure, I haven’t tried this service myself, but if you choose to give it a try I would recommend thoroughly testing your GPT to ensure that the ads you show are not too annoying or interfering negatively with the user experience. Otherwise, the users you have might move elsewhere.

‍

Pricing

AdIntelli doesn’t list any pricing on their websites. However, most ad networks don’t charge publishers a fee. Rather, they take a commission from the revenue from the ads that are served through your platform. I imagine the same being the case with AdIntelli, but please check the conditions with the company directly before integrating their services.

‍