The App Was The Easy Part: A Brutal Guide To Monetising What You Vibe-Coded

There is a moment, usually around day three of a vibe-coding project, when the app works. The buttons click. The database databases. The login logs in. You lean back, look at the thing you conjured out of prompts and spite, and a thought arrives fully formed, like a leaflet through the door: "I could sell this."

You could. People do. But before you start mentally spending the recurring revenue, let us walk through what actually happens next, with numbers, because the gap between "I built an app" and "I have a business" is where most AI-generated dreams go to be composted.

First, the uncomfortable truth: nobody cares how you built it

This cuts both ways, so let's be fair about it. Your customers do not care that you built it in Lovable over a weekend instead of hiring a dev team for six months. They also do not care that you built it in Lovable over a weekend. The tool is invisible to the buyer. Nobody has ever paid £9 a month for an invoicing app because they admired the framework.

This means AI has not changed the monetisation problem at all. It has changed the production problem, dramatically, wonderfully, and in doing so it has made the monetisation problem worse, because now everyone can build the thing. When the cost of making software collapses, the value of having made software collapses with it. What retains value is everything that was always hard: finding people with a problem, proving you solve it, reaching them, charging them, and keeping them. The App Store gold rush taught this lesson in 2010. The AI gold rush is teaching it again, to a new class of people holding a new shape of shovel.

Problem one: does anyone actually need this?

CB Insights has been collecting startup post-mortems for years, and the most famous finding from that graveyard is that "no market need" tops the list of reasons startups die, cited in 42% of the failures they analysed. Not "ran out of money". Not "bad code". Not "the AI hallucinated the entire backend". No market need. They built something nobody wanted, elegantly, and then acted surprised.

And here is the trap specific to AI-built apps: the build is so fast and so cheap that it skips the part where reality gets a vote. A traditional founder had to spend six months and real money before launch, which forced at least some conversation with potential customers along the way, if only to justify the invoice. You can now go from idea to deployed product without speaking to a single human being. This is not a feature. It is a hazard. The app works perfectly and solves a problem that exists primarily in your head.

Validation is not asking your friends. Your friends lie to you, lovingly. As we said in our piece on making money with Lovable sites, the Friends and Family Trap is real: "That's a good idea" is a social reflex, not market research. Validation is a stranger giving you money, or at minimum a stranger giving you an email address and a credit-card-shaped expression of intent. Until that happens, you do not have a product. You have a demo.

Problem two: the numbers are genuinely grim

Let us ruin your weekend properly. RevenueCat's 2025 State of Subscription Apps report, drawn from more than 75,000 apps and over a billion transactions, found that nearly 20% of subscription apps reach $1,000 in monthly revenue. Which is another way of saying 80% don't. Five per cent reach $10,000 a month. And when they compared newly launched apps, the top 5% generated more than 400 times the first-year revenue of the bottom 25%, $8,888 versus a maximum of $19.

Read that last figure again. The bottom quartile of new subscription apps make, at most, nineteen dollars in their first year. Not per month. Per year. That is not a business, that is a rounding error with a landing page.

And even if you get subscribers, keeping them is its own knife fight: nearly 30% of annual subscriptions in the same dataset are cancelled within the first month. People subscribe to your app the way they join a gym in January, and with the same follow-through.

Problem three: pricing, or 'what would I pay?' is not research

Every indie developer prices their app the same way: they ask themselves what they would pay, remember that they are a cheapskate who runs everything on free tiers, and charge $4 a month. Then they wonder why the maths never works.

The founder of EventCatalog wrote a wonderfully honest account of exactly this disease. He priced against what he, a developer, would personally pay, benchmarked against free and open-source tools, and completely missed that business buyers evaluate the cost of the problem, not the hours of code behind the product. A tool that saves a company ten hours a month is not competing with "free". It is competing with the salary cost of those ten hours. Price accordingly, or discover that fifty customers at $4 a month is $200 a month, which does not cover your AI token bill, let alone your time.

Pricing is not a number you divine in the shower. It is a hypothesis you test against actual purchase behaviour. Stated interest is worthless; people will tell you £20 a month is reasonable and then balk at £5. The only honest pricing signal is someone paying, or not paying, at a real checkout page. Everything before that is theatre.

Problem four: the cost of sales, the bit nobody budgets for

Here is where the weekend app meets the weekday reality. Building the app cost you some AI credits and a few evenings. Acquiring a customer costs actual money, forever.

• →Paid acquisition: average cost per install in 2025 ran around $4.70 on iOS and $3.70 on Android. That is an install, not a customer. If your free-to-paid conversion is the industry-plausible 2 to 5%, each paying customer cost you somewhere between $75 and $235 in ad spend before they pay you a penny. At $5 a month, you need them to stay subscribed for well over a year just to break even on the advert.
• →Free-to-paid conversion: across more than a thousand products surveyed, 2 to 5% is the realistic band. That means 95 to 98% of everyone who tries your thing will never pay. Your funnel is not a funnel. It is a sieve.
• →Content and SEO: the 'free' alternative, which costs you six to twelve months of writing before it returns anything, as anyone who has ever run a blog knows. Ask us how we know.
• →Your time: the most expensive input of all, and the one nobody invoices themselves for. Every support email, every refund, every 'quick feature request' from a $4 customer is margin leaking out of the bottom of the boat.

This is the cost of sales, and it is why "I'll just put it on the internet and see" fails. Putting something on the internet is not distribution. It is shelving your product in a warehouse the size of a planet, in the dark, behind a door marked with someone else's brand. Distribution is the business. The app is inventory.

Problem five: the AI app has a special extra tax

If your app is an AI wrapper, and most vibe-coded AI apps are, you have a problem traditional SaaS never had: your cost of goods sold scales with usage, and your usage is priced by your supplier. You charge a flat $10 a month; your heaviest user burns $14 of tokens. Congratulations, your best customer is your worst customer.

This is not theoretical. Forefront Chat shut down its free GPT-4 and Claude access because it could not build sustainable revenue against API costs, and it is far from alone in the AI cemetery. The wrappers that survived the 2022 to 2024 cull were the ones that added something beyond a prompt and a pretty interface: proprietary data, embedded workflows, network effects. A thin wrapper over someone else's model is not a moat. It is a toll booth on a road the toll-road owner can pave over at will, and did, repeatedly, every time OpenAI or Anthropic shipped a feature that used to be someone's startup.

To be fair, the picture is not uniformly bleak: RevenueCat found AI apps generated more than $0.63 per install after 60 days, roughly double the overall median. AI apps can make money. The ones that do are differentiated. The ones that don't are a system prompt with a Stripe account.

Your USP: real, or imagined?

Which brings us to the question that decides everything, the one we keep asking on this site because nobody wants to answer it: what is your unique selling point, and is it real?

"It's like X but built with AI" is not a USP. "It's cheaper" is not a USP, it is a race to the bottom you will lose to someone with more funding or fewer bills. "Mine has a nicer interface" is not a USP when everyone else's interface was also generated this morning. A real USP is something a specific group of people cannot easily get elsewhere: your domain expertise, your audience, your data, your integration into a workflow they already depend on, your willingness to serve a niche too small and weird for venture capital to bother with.

And then look at your competition with both eyes open. Not just the other indie apps, but the free tier of the incumbent, the spreadsheet your target customer is currently using, and the option of doing nothing at all. "Do nothing" is the competitor that beats most products. If your app saves someone twenty minutes a month, doing nothing wins.

So is it hopeless? No. It is just not the bit AI does for you.

Here is the honest summary. AI has collapsed the cost of building to nearly zero, and that is genuinely wonderful. It has not touched the cost of selling, because selling is made of trust, attention and distribution, and those are still scarce. The businesses that make money from AI-built apps are the ones that treat the app as the start of the work, not the end of it: they validate with real payments before building features, they price against the value of the problem, they pick a niche they actually understand, they budget for acquisition like it is a real cost, because it is, and they build something on top of the AI that the AI cannot replicate for the next person with a prompt.

The app was the easy part. It was also, if we are honest, the fun part, which is precisely why everyone does it and then stops. The boring, slow, human work of finding twenty people who will actually pay is where the 80% fall away. If you are still standing after that, you have something worth publishing. And if you want the companion piece on whether the platform itself is the problem, we have already answered that one: it isn't. It never was.

The Monday checklist

• →Before building anything else, get one stranger to pay, or at least commit to paying, for the thing you have already built.
• →Write down your USP in one sentence without using the words 'AI', 'easy', 'fast' or 'affordable'. If you can't, you don't have one yet.
• →Do the acquisition maths: your price, times realistic conversion, minus what a customer costs to acquire. If the number is negative, the app is a hobby, and hobbies are fine, as long as you know that's what it is.
• →Name your real competitor, including 'doing nothing'. Then explain, out loud, why you beat it.
• →Set a price based on the cost of the problem you solve, not on what you, a person who builds apps for free in your spare time, would pay.