Project guide

AI mobile app

All 9 cases list app-store search as a channel and 8 charge a subscription; the one grade A case makes 151 dollars a month, the largest self-reported figure is about 7 million dollars a year

Compiled from the 9 cases in our AI mobile app project: 1 grade A (a listing vetted by the Microns marketplace) and 8 grade B (founders' own accounts in interviews, press reports or their own pages). Grade B figures have no platform data to check them against, so the ranges below only show the scale these cases report for themselves.

Compiled 2026-10-23

Key facts

9Cases graded A or B

0Grade C leads

Income ranges use cases graded A or B (grade B is the founder's own account, with no platform data to check it), in USD; grade C leads are listed but not counted

See the steps

How the cases did it, step by step

Each step names the cases it comes from. Public sources give no reliable time data, so there are no weeks or hours.

Step 1 Start from one small everyday jobEach app solves one thing: restoring old photos, logging meals, tracking spending, scoring football technique, logging GLP-1 and peptide doses, turning recordings into transcripts and summaries, or controlling Codex from a phone.Cases: AI old-photo restoration iOS app · Zhao Chunxiang · Florian Vates · Footly · Pep AI · Wave AI · Remodex
Step 2 Get the first version built with AI coding or freelancersRemodex's first version was written overnight with Codex; Wave's founder taught himself to code with ChatGPT as a pair programmer; Flibbo's founder cannot code and hired developers on Upwork; Chen Yunfei built Kitten Fill Light in an hour.Cases: Remodex · Wave AI · Flibbo · Chen Yunfei
Step 3 Let existing models do the core workThe two cases whose sources name the model both call existing services: Flibbo generates video, images and music with third-party models such as Veo 3 and Kling, and Pep AI's meal scanner uses the FatSecret API. The other sources do not say which model is used.Cases: Flibbo · Pep AI
Step 4 Find the first users around launchPep AI built a waitlist in Reddit's weight-loss and biohacking communities before launch; Remodex grew from a Product Hunt launch and a MacStories review; Footly worked with mid-sized football creators rather than big influencers; MonAi was built in public on X and then partnered with a content creator; Flibbo launched to its founder's 900,000 Instagram followers.Cases: Pep AI · Remodex · Footly · Florian Vates · Flibbo
Step 5 Charge a subscription and keep being found in the app storeEight of the nine charge a subscription. Most of Wave's income is renewals, with 800–1,000 installs a day and about one install in ten paying; about three-quarters of the old-photo app's revenue came from App Store search; Zhao Chunxiang charged from day one.Cases: Wave AI · AI old-photo restoration iOS app · Zhao Chunxiang

At a glance

Ranges are the 25th–75th percentile of cases graded A or B, computed by script from the case data; both ends of a ranged figure are counted

Cases graded A or B: 9
Monthly income (annual figures divided by 12): $25,000 – $60,000
Cumulative income (cases with only a running total): $10,500 – $11,500

Skills involved

How many of the 9 cases graded A or B list each skill

Mobile app development6 of 9 cases

Six cases list mobile app development. Footly rebuilt its onboarding three times, treating retention as a product problem.

AI-assisted coding4 of 9 cases

Four used AI coding: Wave's founder taught himself with ChatGPT as a pair programmer, Remodex's first version was written overnight with Codex, and Chen Yunfei, who says he cannot code, built nine products in a year.

Short-video content & ads5 of 9 cases

Five grew through social media and creators: Flibbo launched to its founder's 900,000 Instagram followers and ran TikTok ads, Footly worked with mid-sized football creators, and Pep AI spent 30,000 dollars on influencers in a month that brought back 60,000.

Pricing2 of 9 cases

Zhao Chunxiang charged from day one with no ad spend; MonAi offers a free trial before a subscription, with family plans for up to six people.

SEO2 of 9 cases

About three-quarters of the old-photo app's revenue came from App Store search.

AI retouching1 of 9 cases

The old-photo app restores, enhances and colourises old photos with AI, for about 1,600 customers.

Key points

How big the income is

The only figure vetted by a platform is the old-photo app's: 151 dollars a month, later sold for 2,000 dollars. The other eight are self-reported: Wave at about 7 million dollars a year, Pep AI at 60,000 dollars a month, Flibbo at about 55,000, MonAi's four apps at 40,000–60,000 a month together, and Footly past 10,000; Zhao Chunxiang made 12,000 dollars in five months, about 2,400 a month.

Only one case has a figure vetted by a marketplace
Most of Chen Yunfei's near-million-yuan year came from corporate AI consulting, not his apps
Annual figures are divided by 12 before comparing

Where the money goes

Wave's founder says roughly half is left after Apple's and Google's cut, and he keeps about a third. Flibbo pays third-party model fees on every generation, so its variable costs are high. Pep AI spent 30,000 dollars on influencers in a month that brought back 60,000. The old-photo app cost about 20 dollars a month to run.

App-store fees
Model APIs paid per use
Ads and influencer campaigns

Dependencies and risks

All of these apps sit on app stores and other companies' models. AI notes are being built into big apps and operating systems, and Wave's revenue has been flat for months; an official OpenAI feature could replace Remodex at any time; about 90% of MonAi's income comes from one app; Pep AI handles health data, and its case does not discuss compliance.

Big platforms build the same feature in
Income concentrated in one app or one channel
Health and money data need privacy and compliance work

Cases graded A or B

See them all in the case library →

BSelf-reportedFlibbo: a Turkish founder who cannot code outsourced an AI content app and says it makes about $55,000 MRR$55,000 MRR · monthly
BSelf-reportedFootly: a teenage founder's AI football-training app from North Macedonia, past $10,000 MRR in its third month≥ $10,000 MRR · monthly
BSelf-reportedPep AI: a University of Oregon student's peptide and GLP-1 tracker with a paid AI meal scanner, $60,000 MRR a few months after launch$60,000 MRR · monthly
BSelf-reportedRemodex: a Naples CS student's overnight iOS remote for Codex, about $10,000 in total revenue a month after launch≥ $10,000 revenue · to date
BSelf-reportedWave AI: a former Uber New York GM's solo, vibe-coded AI note taker at about 7 million dollars a year$7,000,000 revenue · annual
BSelf-reportedFlorian Vates: a German developer's AI expense tracker MonAi, $40,000–$60,000 a month across his apps$40,000–$60,000 revenue · monthly
BSelf-reportedChen Yunfei: built the “Kitten Fill Light” app in an hour and hit the iOS charts, but most of his near-million-yuan year came from corporate AI consulting≤ ¥1,000,000 revenue · annual
AData-backedAI old-photo restoration iOS app: about two thousand dollars in its first year, sold for USD 2,000$151 MRR · monthly
BSelf-reportedZhao Chunxiang: three AI apps built alone, including the food diary Bellybook (胃之书), no ad spend, 12,000 dollars in five months$12,000 revenue · to date

Tools

Tools and platforms the cases mention. Every link goes to the official site, and we earn nothing from them.

Tool	Used for	Link
OpenAI Codex	Remodex's first version, written overnight	Official site ↗
ChatGPT	Wave's founder used it as a pair programmer	Official site ↗
Veo 3	One of the third-party models Flibbo uses for video	Official site ↗
Kling	One of the third-party models Flibbo uses for video	Official site ↗
FatSecret Platform API	Pep AI's meal scanner	Official site ↗
Product Hunt	Where Remodex launched	Official site ↗
Microns	The marketplace where the old-photo app was sold	Official site ↗

Platform rules and compliance

Follow Apple's App Store Review Guidelines and Google Play's developer policies; state a subscription's price, period and how to cancel.
Expense, food and health apps collect sensitive personal data: publish a privacy policy and follow local law; Pep AI's case does not discuss compliance.
Follow the terms of every third-party model and API you call (Veo, Kling, FatSecret and so on), and mind copyright and likeness rights in generated images and video.
Sponsored posts by influencers must be labelled as ads under the platform's rules and local advertising law.

Compare

Compare: AI micro-SaaS (small web tools)

FAQ

How much can an AI mobile app make?

The only platform-vetted case, the old-photo app, makes 151 dollars a month and sold for 2,000 dollars; the other eight are self-reported and range from 12,000 dollars in five months to about 7 million dollars a year.

Do I need to code?

Not necessarily. Flibbo's founder cannot code and hired Upwork developers; Wave's founder taught himself with ChatGPT as a pair programmer; Remodex's first version was built overnight with Codex.

Where do the first users come from?

All nine list app-store search. Beyond that: a waitlist in Reddit communities (Pep AI), a Product Hunt launch and a MacStories review (Remodex), mid-sized football creators (Footly), a creator partnership (MonAi), and the founder's 900,000-follower Instagram (Flibbo).

How much do the app stores take?

Wave's founder says roughly half is left after Apple's and Google's cut, and he keeps about a third.

What is the biggest risk?

Big platforms building the same feature into the system or an official app: AI notes are moving into big apps and operating systems, and an official OpenAI feature could replace Remodex.

Want more examples like these?

See them all in the case library →