Blog/Launch

What we learned shipping an AI learning app to both stores

Screenshots, review questions, localisation and the numbers we actually watch in week one.

By The Daniel (Daniel Szalai)

2026-05-26 · 7 min read

X LinkedIn Facebook

The listing is the product for ten seconds

Before anyone opens the app they see an icon, a title, a subtitle and two screenshots. That is the entire product for the first ten seconds, and most of the download decision happens there.

We rewrote the subtitle four times and changed nothing in the app. The subtitle that named the time commitment beat every subtitle that named features.

Screenshots do the selling

Treat them as a poster sequence, not a gallery of your interface.

01One promise per slide A headline you could read from across the room, then the screen underneath it.

02First two carry the weight Most people never swipe past them.

03Vary the backgrounds Adjacent slides in the same colour blur into one image in the gallery.

What review actually asked

Less about the model, more about the money and the words around it.

01Subscription clarity Price, period and renewal must be visible on the paywall itself.

02Restore purchases Missing it is an instant rejection, and it is ten minutes of work.

03Account deletion If you can sign up in the app, you must be able to delete in the app.

Related track

Learn this properly in five minutes a day

This article is the summary. The app is the practice: short lessons, a quiz after each one, and a glossary you build as you read.

Get the app

Localisation is not translation

A translated store listing gets you indexed. A localised one gets you installs, because the search terms people type are not the words a dictionary returns.

We keep every screenshot string in one file per language, so a new market is a translation job and not a design job. The layout was built for three short headline lines, which is the constraint every translator gets told about up front.

Design the screenshot for the longest language you plan to ship, not the shortest.

The numbers we watch in week one

Not downloads. Downloads tell you the listing worked and nothing about the product.

We watch the share of new users who finish one lesson on day one, and the share who come back on day two. If those two hold, everything else is a marketing problem we can solve later.