10 AI earnings stories: how prompts helped build products that make money

Making money from an idea with AI is possible. But documented stories show more than a short prompt: people chose a problem, built several versions, found customers and spent money on growth. Here are 10 stories with sources and an explanation of what each figure actually shows.

How we read the numbers

Revenue is money from sales before expenses, not net profit. MRR is approximate monthly recurring subscription revenue, not necessarily cash collected during a particular month. Stripe or RevenueCat figures on TrustMRR are checked against a connected source, but are not a full business audit. A screenshot or founder statement is weaker evidence. The figures below are dated snapshots and may have changed.

We did not find evidence that these products were created from a single simple prompt. When founders did not publish their exact wording, we do not invent it. The two prompts at the end are our teaching examples, not the founders’ commands.

1. Payout: an app for class-action settlements

American founder Connor Burd launched a service that helps people discover class-action settlements. In October 2025, Shipaton organizer RevenueCat reported $30,017 in revenue, over 17,000 users and 1,750 paid subscribers. That was the result at the competition checkpoint, not a single day's earnings.

RevenueCat says the design, code and assets were produced with Claude Code and Cursor without manually writing code. The product was then promoted on social media: the organizer reports over half a million impressions. Its appeal was a concrete outcome: helping people find payments they might qualify for.

RevenueCat: Shipaton results ↗

2. AbMaxx: an 18-year-old beginner and a focused fitness niche

Lino Leiden had no coding experience before his first app. He started AbMaxx in Lovable: users enter their details and receive a personalized ab-training plan. TrustMRR showed $46,251 in revenue for the preceding 30 days in a snapshot checked on October 24, 2026, using RevenueCat data.

It was not a one-prompt project: Lino spent two months on the first version and then rebuilt it. To reach customers he tested ads and creator content, and offered one partner equity. Advertising costs and product work were part of the story; revenue is not net profit.

Lino's interview ↗ · TrustMRR data ↗ · Recorded snapshot ↗

3. GETebook.ai: a service for creating ebooks

Romanian founder Bogdan Dragomir described himself as non-technical. His GETebook.ai helps turn an idea into an ebook with chapters and a cover. A TrustMRR snapshot on October 21, 2026 showed $35,469 over the preceding 30 days, checked against Stripe data.

The founder said he used Claude Code for the platform and his own analytics. He sought customers through advertising, social media and search. The figure belongs to the service; it does not show that buyers of its ebooks earned that amount.

Case report and verification method ↗ · TrustMRR data ↗

4. 1ClickWebsite.ai: websites for agencies

Bennett Black created a tool that helps agencies make WordPress sites for local businesses. In August 2025 he said he had built it with Cursor and Claude and reached about $1,600 a month. A TrustMRR snapshot on October 24, 2026 showed $15,006 in 30-day revenue, checked through Stripe.

His customers were not merely experimenting with AI: they created websites for their own clients. The founder shared examples on X. Public sources do not establish his take-home profit after costs or reveal his first prompt.

Story and data ↗ · TrustMRR profile ↗

5. VectoSolve: a former landscaper and SVG conversion

Rob from France worked in landscaping and later at a small digital agency. He needed to turn logos into vectors himself. That led to VectoSolve, which converts images into SVG and other formats. A snapshot dated October 20, 2026 showed $2,695 over 30 days on TrustMRR, including subscriptions and credit packs.

In an interview he said he explored ideas with Manus, wrote specifications and refined the product with Claude Code. Search pages and articles helped attract customers. The sum is smaller than some viral examples, but it represents a specific useful service and data checked against Stripe.

Case report ↗ · TrustMRR data ↗

6. GeoSports: a browser game about sports geography

Frank Michael Smith made the first version of GeoSports with Claude over a weekend: players locate places on a map from sports clues. In September 2026 he reported about $40,000 in monthly revenue, mostly from ads. The figure comes from his interview and displayed analytics; we have not seen an independent audit.

There is important context: Frank already had a large sports-content audience, and a team worked on the project after launch. He did more than send a short sentence: he described the concept, requested a specification and spent hours fixing issues. Shareable game results helped bring in new players.

Interview and context ↗

7. 3AK: runners built a coach for runners

Student athletes Christian and Braylin, without coding backgrounds, described their track-and-field app in Rork, with Claude helping them fix issues. In June 2026 they reported roughly 70,000 in total over about six months. The host described dollars, but the founders' quoted statement does not specify the currency; the amount has not been independently verified.

They already had hundreds of thousands of followers watching their sports videos, giving them an audience at launch. The useful lesson is that they knew a specific community's needs: training plans, progress tracking and technique analysis.

Interview and figures ↗

8. Wrestle AI: match analysis for wrestlers

Student George Lampropoulos created Wrestle AI, which analyzes match videos and suggests exercises. In an interview published in December 2025 he reported 17,000 a month within six months of launch. The case report labels the figure as shown on a screenshot, leaves the currency unspecified and does not independently verify it.

He used Rork and ChatGPT but hired a freelancer to add the paywall. His cofounder already had a wrestling audience and helped drive preorders. The example shows how knowledge of a niche and a distribution partner complemented the AI tools.

Wrestle AI case report ↗

9. ToneAdapt: a guitarist made a sound-setting assistant

Kyan wanted to find amp and pedal settings for a particular song. That led to ToneAdapt. In June 2026 the founder showed over 25,000 in revenue over four weeks across the website and mobile app combined. The cited statement does not specify a currency, and a screenshot is not an independent audit.

He described his first app as built using Cursor and Claude Code; the first version took about a week. He then regularly made short videos, tested them and promoted the formats that worked. His own hobby supplied the problem to solve.

Founder interview ↗ · Revenue context ↗

10. Fly: a browser flight simulator and advertising

Pieter Levels built a browser flight simulator using Cursor and Claude models. In March 2025 he reported $87,000 in monthly recurring revenue 17 days after launch. This is the founder's own claim, not independently verified profit.

The game was free to play; revenue came in part from advertising inside the game world and paid features. Levels was an experienced entrepreneur with a large audience before the game. He later described the project as having made money but not becoming a sustainable business.

Claim and date ↗ · Context on sustainability ↗

What these stories have in common

A successful idea usually solved a specific problem: finding a settlement, dialing in a guitar sound or building a client website faster. AI helped make the first version, while sales came from audiences, search, partners or advertising. The founders’ existing audiences and budgets varied considerably.

Do not use someone else’s revenue as your financial plan. Ask early users to try the product, learn what they would pay for, and calculate AI, hosting, advertising, platform fees and taxes. For a game, start by finding real players before forecasting advertising revenue.

Two prompts to test your own idea

These are editorial examples to help you start. They are not the original prompts used by the founders.

I work in [field]. People often spend time on [specific problem]. Suggest three simple solutions that can be tested in a week. For each one, explain who would pay, how to find the first five users and what costs to expect. Do not invent statistics.

Build a prototype of [product description] for [specific audience]. The first version should do only [one action]. Explain how to launch it, test it with five people and fix errors. Do not connect payments or collect personal data yet.

Want to try making your own game? Start with the practical guide and show the result to the community.

How to make your first game with AI → · How AI-assisted games make money → · Telegram ↗