Design your app screenshots for app store search, not for the detail page.
Apple shows the first one to three screenshots in search results when a listing has no app preview. That row is where most installs are decided, so app screenshots for app store listings have to work as an advert before they work as a gallery. Pixel dimensions are the easy half of this job and they are already written down in App Store screenshot sizes. This page is about what goes inside the frames.
What follows: how to choose the first three, what belongs in each frame, what the App Store screenshot requirements actually are now rather than three years ago, and what it costs to localise a set properly.
Plan the first three framesThe short version
A screenshot set is an argument, and frame one is the claim.
Making app screenshots is not documentation work. The set has a couple of seconds to say what the app does for this person and why it beats what they do now. Frame one states the claim. Every frame after it either supports that claim or earns its place with a second reason to install.
Most weak sets fail the same two ways. They open with a logo or a sign-in screen, and they show empty screens, because whoever took the captures used a fresh account with no data in it. Both are fixable in an afternoon, and fixing them is worth more than any amount of work on gradients and device frames.
The first three frames do most of the work
Apple is direct about this. Depending on the orientation of your screenshots, the first one to three images appear in search results when no app preview is available, so those images have to carry the essence of the app. Each screenshot after them should focus on one main benefit or feature. You can show up to ten on the product page, and most apps have nowhere near ten things worth saying.
So order the set by what a stranger needs, not by the order the screens appear in the app. Frame one is the single job the app does, in use, with real content on screen. Frame two is the moment it beats the way they do it today: the receipt that scans itself, the shift somebody else picked up, the bill already split. Frame three is proof that the thing keeps working after day one: a history, a total, a streak, a result that only exists once the app has been used for a week.
There is a floor under this from the review side. Guideline 2.3.3 says screenshots should show the app in use, and not merely the title art, login page or splash screen. A splash frame is not only weak marketing, it is the sort of thing that comes back from review. The same search row carries the icon that sits beside them, doing the identical job in miniature, and it deserves its own pass.
Try it
Which frames land in the search row
Turn on the frames you plan to ship. Only the first one to three of them show up in search results when the listing has no app preview.
Logo on a splash screen, The main job, being done, Sign in and onboarding
2 of those 3 frames show chrome rather than the app. Splash, sign in and settings belong late in the set, or nowhere.
What actually goes inside a frame
Fill the app before you shoot it. Three tasks in a planner looks abandoned. A realistic week in a planner looks used. Populate a demo account with the names, amounts, dates and photos a real customer would have, then capture from that account. This is also the moment when the screens you drew as a mobile app wireframe either hold up or admit that your main screen has nothing on it worth photographing.
Captions are allowed, and they change the whole set. Guideline 2.3.3 permits text and image overlays, and a caption is what turns a picture of a screen into a claim. One idea per frame. Six words or fewer. Put it at the top, where the crop in the search row cannot eat it, and set it large enough to read at thumbnail size, which is smaller than you think. Split a bill in two taps beats Our app makes bill splitting easy, every single time.
Capture from the build you are shipping. A set taken from a version two releases old drifts away from the app it describes, and metadata that matches the app is a review requirement rather than a nicety. If testers already have the current build through ios testflight, take the captures from that same build, and keep the raw files in the repo so the next set starts from something instead of from nothing.
The App Store screenshot requirements, as they stand
The required list is much shorter than most guides make it. Apple requires screenshots for the 6.9-inch iPhone display. The 6.5-inch entry is required only if your app runs on iPhone and screenshots for the 6.9-inch display are not provided. iPad has one requirement, a 13-inch display set, and only if the app runs on iPad. Mac, Apple TV, Apple Watch and Apple Vision Pro carry their own required entries if you ship to those platforms.
Sizes you leave empty are filled by scaling the larger set down, which is why a single good iPhone set covers the iPhone slots. You can upload 1 to 10 screenshots per size, per localisation, as .jpg, .jpeg or .png. Images cannot include alpha channels or transparencies, so flatten anything exported from a design tool on a transparent background before you upload it.
Screenshot sizes for the App Store do change, and older articles still ask for five or six separate iPhone sets that the current specification does not list. Read the specification page before you build an export pipeline around a list, because the pipeline is the expensive thing to redo.
Localising the set, then finding out if it worked
Apple asks you to localise the description, keywords, app previews and screenshots for each market where the app is offered, and every localisation carries its own set of up to ten images. Translating the captions is the visible half of that. The invisible half sits inside the screens: currency, date format, example names and addresses, and string length, since translated text often runs longer than the English it replaced and quietly breaks a layout that was tuned by eye.
This is the argument for producing screenshots with a repeatable capture rather than a folder of hand-edited files. When the caption layer, the device frame and the language are separate inputs, a new language is an afternoon. When every frame is a flattened export, a new language is the entire job again. Decide early which markets earn a real set, because a localised set for a country with no installs is expensive decoration.
Then find out whether the set works instead of arguing about it. Product Page Optimization in App Store Connect runs up to three alternative treatments against your current page, for up to 90 days or until you stop it, on a share of traffic you choose. App Analytics reports impressions, conversion rate, improvement and a confidence level for each treatment, and Apple suggests waiting for high confidence before adopting one. Custom product pages are the other half of the same idea: separate versions of the page with their own screenshots and promotional text, each on its own URL, so a campaign can land on a set built for it.
What each way of making the set gives you
| How the set is made | Shows the real app | Caption per frame | Redo after a UI change | Extra languages |
|---|---|---|---|---|
| Raw simulator captures | Yes | No | recapture by hand | recapture per language |
| Captures plus a design file | Yes | Yes | half a day of editing | one design file per language |
| Screenshot generator service | only what you upload | Yes | re-upload and re-render | paste in the strings |
| A capture script in your repo | Yes | from the script | re-run the script | re-run per locale |
| Agent capture inside an app builder | Yes | Yes | ask for another pass | if you supply the strings |
Making the set, and keeping it honest
The cost is never the first set. It is the fourth. A screen changes, a caption stops being true, one language goes stale, and the listing drifts away from the app it describes. Anything that makes recapturing cheap is worth more than anything that makes the first set prettier. The useful question is not which design tool to buy, it is what happens on the day the main screen changes.
Newly is an AI app builder: you describe the app in plain English, it writes a real React Native and Expo project you own, and it runs that app on a cloud iPhone or Android simulator while it builds. The simulator has a save screenshot button, and you can ask the agent to capture the running app and save the images into the project, so the raw frames come from the real build rather than a mock-up. On a Mac local chat the agent can also produce styled store sets, though AI-generated backgrounds need your own OpenAI API key and fall back to a plain gradient without one. Plans are $25 a month, there is no free plan, and uploading to TestFlight still needs your own Apple Developer account.
None of that decides what the frames say. A tool can put a caption on a gradient. It cannot tell you which of your features a stranger cares about, and it will happily produce ten beautiful pictures of the wrong thing. That part is the ordering exercise above, done with two or three people who have never seen the app.
Questions people ask about App Store screenshots
One is the technical minimum: a single 6.9-inch iPhone screenshot, plus a 13-inch iPad one if the app runs on iPad. You can upload 1 to 10 per size, per localisation. In practice, ship as many frames as you have real things to say and stop there. Five strong frames beat ten padded ones.
Build the app, then shoot the frames that sell it
Describe the app you want, watch it run on a cloud simulator, and take the screenshots from real screens with real data in them.
Start building