Hundreds of museum photos sitting in your camera roll, and someone will pay $40 a year to sort them: artscan's playbook

A small team in New York built an app that does one thing: pulls the exhibit photos you've taken out of your phone, identifies which artwork each one is, and sorts them by museum and map. Google has offered identification for free for ten years, so this one sells "organization," at $39.99 a year. It ran on iOS for a year, then in October launched an Android version, switched its existing Facebook ads over to it, and gained about 1,400 installs in a little over a week. Below: what it actually sells, where the traffic comes from, who finds it expensive, and whether you can use this playbook for something else.

A New Yorker and the photos he took at the Met

In 2026 Oct 16, a US user gave artscan five stars on the App Store: "Just added the photos I took recently at the Met. I'm from New York and go to museums often... It saved all the paintings and exhibits from my phone and sorted them, and even the few I couldn't recognize myself it got right."

Scrolling further back, in Jul 8 someone wrote "I've been waiting years for this app," and in Jun 27 another wrote "I've been hoping for an app like this" and then gave two stars, because she uploaded 100 photos and "the AI got almost all of them wrong."

That's the whole story of artscan: people who want it really want it, and whether it does the job well is a separate question.

What it is

artscan: Museum Photos, developer ArtScan AI LLC, New York. The iOS version has been on the App Store since 2025-10-13; in the US store it has 11-06 ratings, averaging 1,028 stars (App Store data, 4.69 11-06). The Android version went live 2026-10-10.

The core feature after opening is called Auto Review: it scans your whole camera roll locally on the phone, picks out photos that look like artwork for you to confirm, then batch-imports the confirmed ones, turning each into a card with the artwork title, artist, and museum; labels next to exhibits can be read from the photo too; the cards are then filed by museum and map. The first line of the Play store page reads "Turn museum visits and photos on your phone into a personal art collection," and the ads say "Stop losing your art photos to your messy camera roll!"

Note that it never competes with Google Lens on who recognizes more accurately. Its pitch is: you've already taken hundreds of these, they're buried in your camera roll, and I'll get them sorted for you.

The package names on both platforms are app.rork.*, meaning it was built with Rork (a tool that generates mobile apps with AI).

Why it suddenly spiked this week

Not word of mouth, ads.

Search "artscan" in the Meta Ad Library: the advertiser page "ArtScan / ArtScan AI" first started running ads on 2026-01-26, creatives kept running from February through September, and in early November about 51 ads are live. Starting in Oct 19 a batch of new ads went up, followed by 10-22, 10-30, 11-02 (7 ads a day), and 11-04, and these new ads' landing links go straight to the Google Play Android page.

Now the Android install counts (Google Play installs, day by day):

The two jumps line up with two batches of new ads. Store install counts only update once every day or two, so they look like steps. On the charts it only appears in the Arts & Design free charts in France, the Netherlands, Australia, the UK, Canada, and Germany, with a best rank between 60 and 120. That category has a low bar; 1,000 installs are enough to chart.

On iOS, 11-02 it simultaneously entered the Taiwan Photo & Video free chart and the top 100 of the top grossing chart, and ratings suddenly gained more than a dozen. My guess is the ad spend expanded to Taiwan, unverified.

I couldn't find any of its content on TikTok, X, in the media, or on Product Hunt. It's simply an app quietly buying traffic through Facebook and Instagram.

Where the money comes from

Every feature requires a subscription: on Android it's listed at $39.99 a year (7-day trial) or $99.99 lifetime; iOS reviews also mention earlier prices of $4.99 and $7.99 a month. After opening, you first answer 6 questions you can't skip, then hit the paywall: the standard subscription-app funnel.

Revenue is not public. A third-party data site estimates about 75k iOS installs and under $10k monthly revenue. This is an estimate.

The high price is what makes buying traffic work: one annual subscriber is enough to cover the ad cost of dozens of installs, and "middle-aged Americans and Europeans who often visit museums" are easy to find on Facebook and have money.

It's also good at collecting good reviews. A one-star review in Oct 25 laid out the flow clearly: "Log in with Apple or Google, upload 3 photos, it runs an 'AI analysis' but you never see the results, then immediately pops up asking you to rate it." On Android, 1,709 installs produced 73 ratings, about 4%, where a typical app gets under 1%. So the score is 4.6–4.7 stars, yet most of the review text is complaints.

What users are complaining about

Two complaints, over and over:

• Expensive, and used too little. "People visit a museum once or twice a year, this app gets used for a few days a year, not worth a subscription" (10-06, one star); "$40 upfront is outrageous, I'd rather pay $5 a month. The idea is good, I really wish I could use it" (11-01, three stars); "An app this simple, and the subscription costs as much as Netflix" (7-27).
• Gets it wrong. Contemporary art and "which museum" are the easiest things to miss: "During the trial I uploaded 3 photos, 3 museum pieces were all wrong, one of them has been hanging at MoMA for nearly 60 years" (6-22); "Auto-scanning the camera roll treats all kinds of non-art photos as art, so you have to swipe left and right on each one" (8-09).

Plenty of people want it, but few will stay at this price and this accuracy. That's both its weakness and an opening for followers.

Who it's competing with

"Recognize a painting" is a red ocean; the slot of "automatically pull out the exhibit photos you've already taken, recognize them, and organize them into a collection" is occupied only by artscan, which is seriously building it and spending money on ads. Copycats have already arrived: Museum Log, launched in August, has store copy nearly lifted word for word ("Turn your museum photos into your personal art logbook"), and designer January's Museum Diary preview video got about 3.2M views on TikTok, but both products have only single-digit ratings. The idea has been copied; the playbook has not.

By the way, there's an IndieHackers post online where the author says she sold her own ArtScan, which she built over 2024 years, on Flippa. That's a different app with the same name (Painting Recognition - ArtScan, launched in August 2024), not the same as this artscan, so don't mix them up (inferred from launch dates and scale; the original post doesn't give the app id).

If it were you

One-line verdict: Worth a try. This playbook, swapping "AI identifies X" for "automatically organize the X already in your camera roll into a collection," with a high-priced subscription, adding Android a few days after the iOS creatives prove out, then switching the same Meta ads over, is clear and repeatable, and few people know about it yet. Its weaknesses are also obvious: people visit museums once or twice a year, so annual fees are hard to renew, and its growth is entirely ad-driven, so it stops when the ads stop.

Who it suits: 1–2 people, one of whom has run Meta ads and can calculate what one paying user costs to acquire; able to ship both platforms in a week or two with tools like Rork or Expo.

Who it doesn't suit: Anyone counting on organic traffic (this category has none, and artscan doesn't either); anyone wanting to compete with Google Lens on recognition accuracy; anyone unwilling to "use some tactics" on paywalls and rating prompts.

How to do it: Two paths.

• Build the same scenario but fix what it does badly: charge per use or per "trip" ($4.99 to unlock the 200 photos from this European trip), give a free allowance generous enough that people see results first; when recognition is wrong, let users photograph the label to correct it instead of pretending it's accurate. Its most common bad reviews are about pricing model and accuracy, and both are ready-made product directions.
• Pick something that already piles up in camera rolls and gets photographed more often: wine labels, restaurant dishes, sneakers, concert tickets, buildings photographed on trips. Before choosing, check two things: do people really snap hundreds of these casually, and is there already a big player (plants have PictureThis, wine has Vivino, so stay away from those two).

How to acquire users: Copy its playbook. Open the ad with the pain point, not the AI ("You've probably lost track of the exhibit photos on your phone"). Use a screen recording of "scan your camera roll, hundreds of photos turn into a card wall." In the US or UK, test 10–20 creatives on a small budget first. Only scale once you've pushed the cost per paying user below half of first-year revenue (inference). On first use, bulk-import the "past" photos so users see the payoff before the paywall. Offer two tiers: annual and lifetime.

When to move: Launch before spring 2027. Travel and museum-going peak in Europe and the US starting at Easter, so test your creatives before then.

Signs you've misread it: ① artscan's Meta ads visibly shrink or stop in early 2027, meaning its own numbers don't work. ② iOS rating growth stalls at single digits per week while prices keep getting cut, meaning the high annual price can't hold. ③ Google Photos or Apple Photos ships a built-in "artwork" album with recognition, and the system feature eats this niche. ④ The reverse: copycats like Museum Log start running ads and their ratings climb, meaning the niche is bigger than it looks and you should move fast.

---

Data and sources: Google Play installs and ratings (daily, 10-29 to 11-06), Play category charts by country; App Store ratings and reviews (US store, 11-06); Meta Ad Library web (search "artscan", currently running); iOS installs and revenue are estimates from third-party data sites, unverified; TikTok and X searches; original IndieHackers post (2026-08-20).