I built a ChatGPT-powered custom travel guide generator and it became profitable 24 hours after launch

Background

I recently paused freelancing (creating visuals for entrepreneurs) because it doesn't fit this stage of my life—mother of 2 young kids, full-time mothering, very limited time in a day, can't leverage my time enough, have so many ideas I want to execute etc.

Back in 2020, right after having my first baby, I attempted my first solo internet product called Apparent. It would have been what I now describe as the Headspace for parents. After spending $1,000 on the voice clips, I decided my approach was wrong.

How I started building things that make money

Shortly after, I enrolled in the Small Bets community and quickly came up with a better way to start making money online. Leveraging my then ~3,000 followers on Twitter, I built a visual design micro agency called VisualNeuron.com, and right after I launched it, I started earning decent amounts of money. My work setup was similar to that of a freelancer. Get orders from entrepreneurs to visualize their ideas in my style.

Then arrived my second baby. I took 6ish months off to focus on my family and by the time I got back, I had received new orders but I didn't feel like going back because of the reasons I mentioned in the first paragraph.

So I made a plan, more context here:

Launching 12 products in 12 months

At first, it felt terrifying and I thought, "If it feels terrifying, it's the right direction."

July 2023. The first product I launched took 20 days to build from start to finish (working a max. of 2-3 hours a day, only on weekdays.) It became profitable the day after launch and has been my most successful launch ever. Today I want to tell you how I built it.

Why ChatGPT

In my founder circles, I had been seeing creative projects making use of the ChatGPT API and although there are many people on my Twitter who look down upon this kind of projects ("ChatGPT wrappers"), I wanted to give it a go.

After all, language models are in their infancy and 90% of the people we see on the streets don't have a clue what ChatGPT is. It's still early. Building something making use of language models would be good practice. Also, I'm (not humble-bragging but) a bit too good at providing instructions in the written format—a bit of an instruction micromanager, you might say. So trying my hand at prompt engineering made sense.

The idea

The idea I was searching for would "provide a gateway to an information pool." I quickly settled on an idea. I'm a big travel nerd so I'd build a custom travel guide generator.

First I wanted the final product to be customized mini ebooks. Think Lonely Planet guides written just for you. I then lost the domain customtravelguides.com to a surprise rival and started searching for another domain.

Lightbulb moment

Around the time, I was debating with my husband, who helps me with the backend, what the quickest development method would be.

First, better if we don't log anything. So no database. Second, everything should be automatic because I have zero time for manual work during the day.

Generating custom ebooks formatted in the way I had in mind would be tough. Simultaneously, I found a new domain name, TutoredTrip.com, and got a lightbulb moment. Can we deliver this "guide" as if it's an email course? With a newsletter-like format? Husband said it would be a lot easier than PDFs. I loved the idea of a ChatGPT-powered email course. This would be fun.

How I built it

I got to work. For a solid two weeks, I worked on my prompts, figuring out how to get the best information by providing different profiles. I experimented to see if the model would make different recommendations when I provided a sportsy person's profile and a foodie's. I crafted a set of prompts. This was the biggest chunk of my work and I'm proud of my quite lengthy prompts.

I then crafted a quiz on Tally by researching what kind of things people care about on a trip. The results of this quiz would be fed into the prompts.

I then crafted email templates on Mailgun with static texts and variables.

I bought the domain. Made a quick landing page on Carrd, hesitated to post it online because it bothered my perfectionist side, got feedback from a few Twitter friends, worked on it and other things for a week more, and finally...

Launch

I launched TutoredTrip.com 2 days ago—posted it on my Twitter and it started selling right away. I haven't tried any other marketing method yet. My launch tweet has been bringing in sales.

The morning after the launch, I woke up to a handful of sales notifications. That's when I realized I had recovered all of the development costs—mostly going to ChatGPT API, as all other tools I used were basically free. So it was profitable!

Business model

Here's how it works: the user fills out a quiz which I tried to make as lightweight and fun as possible and so far it's gotten nothing but positive feedback. After the quiz is submitted, the person's guide is generated right away and scheduled to be delivered in 5 emails over the course of 4 days.

My laser-focused prompts are fed into variables in a newsletter-like flow of information that I designed for maximum enjoyment. It's not only travel recommendations, there are jokes, language tips, trivia questions, etc. scattered across.

ChatGPT and other language models get a bad rep for delivering false information but I think people are missing the point. These models are going to get better and better and their references are going to get a lot more accurate. That's why experimenting with methods of information delivery matters more than the quality of the information they deliver today. And this project does just that.

I forbade it to recommend specific restaurants, stores etc. because that's where errors can happen. It's just not the value here. The value here is to instantly get an overall sentiment of the city, customized for what you're looking for on trip. ChatGPT is great with general knowledge and it's great with categorizing it for specific personas. I've run numerous tests and am satisfied with the majority of the results.

Lastly, customers get a heads up about being possibly given false information so important things should be double-checked before the trip—truthfully, I double-check everything I read on blogs too, so it's fair.

Going forward, each tutorship will cost me $0.40. I've priced each round of tutorship at $5, which is quite affordable for the user and which I'm happy with too.

It leaves enough margin for (potential) marketing spending and if I can tap into a more lucrative market, I can always up the value and the price.

I think the success of this launch partially relates to putting an affordable price tag, which is perfectly fine. I believe in leaving money on the table.

I imagine TutoredTrip as a marketing tool for airlines, hotel chains, booking.com, Airbnb, or any large tourism operator that wants to deliver their customers a fun, customized travel guide experience.

As a B2C product, it could work great at the top of the funnel for bloggers, travel influencers, etc. With larger marketing budgets and with strict fact-checking, it could disrupt travel guide publishers like Lonely Planet.

For now, I'm invested in delivering a product customers will have fun with so my main focus is to collect feedback from paying customers, try my hand at marketing (which I'm not great at!) ,and improve my prompts.

Thanks for reading thus far. Happy to answer any questions you might have!