Subscribe
AI for App GrowthFacts as of

Why AI Apps Earn More but Keep Less (And How To Fix It)

The short answer

How do AI apps make money? Revenue, pricing and retention in 2026

In RevenueCat's 2026 data, the typical AI app makes about 41% more per paying user in its first year (around $30, against $21 for other apps). But its subscriptions last the year less often: on annual plans, about 21 in 100 are still active after 12 months, against about 31. The fixes are about the plan, the promise and the habit, not the AI.

Key facts

Chapters

  1. 0:00AI app revenue: 41% more per paying user
  2. 0:36What counts as an AI app
  3. 1:12AI app pricing: monthly vs yearly plans
  4. 2:52AI app free trials and conversion
  5. 4:35AI app revenue per user: month one vs year one
  6. 6:04AI app refunds and monetization mix
  7. 7:43AI app retention: the catch
  8. 9:13What this data can't tell you
  9. 10:103 changes for an AI app

Transcript

Edited for reading

Teal is a fictional AI photo editor; its screens are illustrative. Every other number comes from a real report, listed under Sources. The comparisons are between different apps, so they show patterns, not promises for your app.

AI app revenue: 41% more per paying userWatch from 0:00

The typical AI app makes about 41% more from each paying user in its first year. Around $30, against $21 for other apps.

But look at who's still around a year later. On annual plans, about 21 of every 100 AI app subscriptions are still active. For other apps, it's about 31.

So which number wins? Five findings will tell you, then three changes to make before you set your prices.

The data comes from RevenueCat's 2026 report, covering more than 115,000 apps. Our example app, Teal, is an AI photo editor we made up. The numbers aren't.

What counts as an AI appWatch from 0:36

First, what counts as an AI app. RevenueCat means apps where AI is the main value. Think chatbots, AI writing tools and AI photo editors.

By that test, about one in four subscription apps is AI-powered. RevenueCat processes the subscriptions for the apps in its data, so these figures come from more than a billion real transactions.

In photo and video apps, it's more than six in ten. That's Teal's category.

The report doesn't say how many AI apps it counted. So every RevenueCat number here describes the typical app in each group. None of it predicts yours.

AI app pricing: monthly vs yearly plansWatch from 1:12

Finding one: what AI apps actually sell. Which plan makes up most of an AI app's subscription sales? Weekly, monthly or yearly?

For apps without AI at their core, yearly plans sell best. They're about four in ten subscriptions sold. AI apps flip that. Almost 60% of the subscriptions they sell are monthly. Other apps? About a quarter.

Yearly and weekly plans both shrink to make room. Other apps lean on weekly plans too: about three in ten of the subscriptions they sell are weekly.

Put it in one picture. Of every ten subscriptions the typical AI app sells, six are monthly. Only about two and a half are yearly.

A yearly subscriber pays once and is committed for twelve months. A monthly subscriber gets a fresh chance to leave every month.

The report doesn't say why. Our take: a lot of AI apps look like they copied the big chatbots' monthly plan.

The better question is how often your users come back. If people open your app every day, a yearly plan is an easy yes. If they show up twice a month, monthly might be the honest offer.

And the mix matters, because monthly plans last a year far less often than annual ones, in both groups. Finding five shows by how much.

Teal, from our video Why Your App Pricing Isn't Working (And How To Fix It), leads with a weekly plan and keeps monthly and annual behind it. That choice came from how people use it. This data doesn't change it. It just says don't let monthly win by default.

Checkpoint for your app: which plan does your paywall push, and did you choose it on purpose?

AI app free trials and conversionWatch from 2:52

Finding two: AI apps are very good at getting a first try.

Look only at apps that offer a free trial. 8.5% of AI app downloads start one within a month. For other apps, it's 5.6%. That's about 50% more for AI.

Our take: AI is easy to be curious about. Curiosity gets the tap. It doesn't pay the bill.

Now look at a second measure: downloads that go on to pay. Within five weeks, 2.4% of AI app downloads have paid. For other apps, it's 2%.

The AI lead drops from about 50% to about a fifth. And this second measure covers all apps, trial or not.

Try it per thousand downloads. Among apps with trials, about 85 AI app downloads start a trial, against 56. Across all apps, about 24 AI app downloads turn into a payer, against about 20.

So AI apps win the curiosity contest by a mile, and the money contest by a nose. Lots of people try an AI app. A smaller share of that interest turns into money.

For Teal, a busy trial screen isn't the goal. The payment after it is, so that's the number Teal should watch. Teal's trial is three days, short enough to make the first edit count.

Checkpoint: split your own funnel the same way, trial starts per thousand downloads, then payers. If the second number lags, the trial is doing the selling and the app isn't.

One warning if you read the report: its summary calls the trial-start number trial-to-paid. There's no AI figure for trials that convert.

Keep that shrinking lead in mind. The number that undoes the +41% comes in finding five.

AI app revenue per user: month one vs year oneWatch from 4:35

Finding three: where the extra money comes from. So far, AI apps sell mostly monthly plans. They get tried more. And the lead shrinks before anyone pays.

How much of the year-one gap is already there after one month? In the first 30 days, the typical AI app makes about $19 from each paying user. Other apps make about $13.60. That's about $5 more per payer, in month one alone.

Compare that dollar gap with the full year's. Roughly 60% of it is already there by day 30. That's our own calculation from two of RevenueCat's charts, so treat it as a rough guide.

After the first month, both groups grow their revenue per payer at about the same rate. So most of the AI premium appears to be won in month one.

For Teal, that puts the weight on the first session. The edit people came for has to land before the first bill does.

Checkpoint: look at what your average payer brings in during month one, then what they add by month twelve. If almost all of it comes in month one, your renewals are the weak spot.

There's a popular idea that AI is where the biggest app money is. Look at the app at the 90th percentile in each group, and the AI edge nearly disappears. So the premium belongs to the typical AI app. Near the top, both groups earn about the same.

Our take: AI apps tend to win the first sale more easily. That alone doesn't make a top earner. And earning more in month one only helps if people are still there in month thirteen.

AI app refunds and monetization mixWatch from 6:04

Finding four: AI apps hand back more money. At the typical AI app, 4.2% of new subscriptions get refunded in their first billing period. That's someone who paid, then got their money back before their first period ended.

At other apps, it's 3.5%. About a fifth more refunds for AI.

Our take: a refund is a promise the app didn't keep. With AI apps, that promise is often the demo.

Per thousand new subscriptions, that's about 42 refunds at the typical AI app, against 35.

AI apps also lean harder on subscriptions alone. About two thirds of them sell nothing else. Fewer offer a lifetime option: about one in five AI apps, against more than one in four other apps.

And credits? RevenueCat doesn't track them on their own. The closest thing, one-off consumable purchases, is only slightly more common in AI apps.

Our take: for an AI app, a lifetime plan is a promise to pay for someone's usage forever. Price it like one, or skip it.

Checkpoint: read your refund reasons from the last month. If they say the app didn't do what the ad showed, fix the ad before you fix the app.

Now set that beside some advice from the same report. Growth adviser Phil Carter writes that serving an AI subscriber is no longer close to free, because language models cost money every time they run. He says many AI apps are pushing users toward annual plans and higher tiers.

Yet in RevenueCat's own data, a far smaller share of what AI apps sell is yearly. The advice is out there. The typical AI app's plan mix doesn't reflect it.

AI app retention: the catchWatch from 7:43

Finding five: the catch. RevenueCat checked which paid subscriptions were still covered twelve months later, one plan length at a time. For an annual plan, that twelve-month check is really one moment: the first renewal.

On annual plans, about 21 of every 100 AI app subscriptions made it. For other apps, about 31.

Start with 1,000 annual subscriptions. A year on, about 211 are still active at the typical AI app. At other apps, about 307. That's close to a hundred fewer renewals per thousand.

On monthly plans, it's about six in a hundred, against more than nine. Per thousand monthly subscriptions, that's about 61 still active after a year, against 95. Weekly plans barely last a year for anyone, AI or not.

So on annual and monthly plans, AI app subscriptions are roughly a third less likely to last the year.

Now remember finding one. AI apps sell mostly monthly plans, and monthly plans rarely last a year in either group. So the +41% per payer is real. It just sits on subscriptions less likely to last a year.

Why do fewer AI app subscriptions last the year? The data doesn't say. Novelty wearing off is one possibility. Category is another.

For Teal, that makes one number worth watching from launch day: of its first hundred annual buyers, how many come back for year two?

Our take: for an AI app, the first renewal is the real launch.

What this data can't tell youWatch from 9:13

A second dataset doesn't fully agree. Adapty compares AI apps with the average of all apps, so it's a different yardstick.

In its data, AI apps start trials at about half the average rate. RevenueCat found the opposite. The samples, the definitions and the comparison all differ, and neither report says which one explains the gap.

Both agree on two things: more revenue from AI apps, and fewer subscriptions kept a year on. Adapty also finds AI apps earn more per install: about $1.44, against 84 cents for the average app.

And some things neither can tell you. Whether AI causes any of these gaps. How many AI apps were counted. What they charge, or what running the AI costs them.

These are medians, the middle app in each group. Your app can sit far from the middle, in either direction.

AI apps also cluster in photo and video and productivity. So some of these gaps may come from the category.

3 changes for an AI appWatch from 10:10

Our take: read these numbers as warning lights for your own dashboard. They're not a forecast for your app.

So here's what we'd change at Teal. Three things, our call, for a made-up app.

One: pick the lead plan on purpose. Teal keeps its weekly plan up front and marks the annual plan as the best value for regulars. Monthly plans kept far fewer subscriptions than annual ones, in both groups. An annual plan the regulars can see is cheap insurance.

Two: before the first charge, show exactly what the plan includes, and send a reminder. Fewer surprises should mean fewer refunds.

Three: give people a reason to come back, long before the first renewal. For Teal, that's a fresh edit idea built from the photos people already have. The first renewal is where finding five bites, so that's where the habit has to be in place.

Notice that none of these is about the AI itself. They're about the plan, the promise and the habit. That's Teal's list. Yours comes from your own numbers.

Should your AI app be buying ads inside ChatGPT? That's our video Why ChatGPT Ads Don't Pay for Most Apps (And How To Fix It).

Sources

  1. RevenueCat, State of Subscription Apps 2026 (2025 data, 115,000+ apps; AI vs non-AI apps, medians; includes commentary by Phil Carter, Elemental Growth, shown as opinion)
  2. Adapty, State of In-App Subscriptions 2026 (2025 data, 16,000+ apps; AI apps vs the all-app average) (5 March 2026)

Caveat: Where RevenueCat's text and charts disagree, we used the chart values: weekly retention is 1.2% vs 1.7% (the summary page's 2.5% vs 3.4% are 75th percentiles), and the summary's "trial-to-paid" figure is the trial start rate. The per-1,000 figures and the day-30 share of the year-one gap are our own calculations from the medians, so treat them as indicative. RevenueCat does not publish how many AI apps it measured.

Watch next