App Store Screenshots That Convert in 2026

The first three screenshots are the whole pitch

Apple's own product-page research puts roughly 60 percent of the install decision on the first three screenshots, and there is a structural reason. On iOS, the first three portrait panels render directly in search results, so a browsing user sees them without ever opening your page. That is a mini-narrative window, and if those three images do not resolve into a complete thought, you lose the person before they tap "View More."

Treat the three like a three-beat story:

1. Panel one, the hook. The single clearest statement of what the app does, over your most representative screen. This frame matters more than the next nine combined.
2. Panel two, the mechanism. How it works, in one glance. Show the core action, not a settings page.
3. Panel three, the payoff. The result the user came for. The finished thing, the saved state, the win.

Problem, solution, outcome. If you can only make three panels great, make these three and stop.

Size every screenshot to the pixel (2026 specs)

App Store Connect validates screenshot dimensions to the exact pixel, and a file that is one pixel off, or that picked up padding from a device frame, gets the "the dimensions of one or more screenshots are wrong" error. The current required sizes as of July 2026:

• 6.9-inch iPhone (17 Pro Max, 16 Pro Max, 15 Pro Max): 1320 by 2868 portrait, or 2868 by 1320 landscape.
• 6.7-inch iPhone (16 Plus, 15 Plus): 1290 by 2796 portrait. A common, accepted alternate.
• 13-inch iPad Pro: 2064 by 2752 portrait. Required only if your app runs on iPad.

You supply the largest size per device family and Apple scales it down for smaller devices, so you are not exporting a dozen variants. Files are PNG or JPEG, RGB color, and must not carry an alpha channel. Google Play is far looser: phone screenshots between 320 and 3840 pixels on any side, a 2:1 or 16:9 aspect band, up to 8 per type. Because Play does not pin you to one exact size, the same rendered artwork usually exports cleanly for both stores if you design at the iPhone spec first and letterbox for Play.

Our own SparkQuest render set lives in three folders, one for Android, one for iPhone, one for iPad, exactly because the iPad frames have to be re-exported at 2064 by 2752 and cannot be scaled from the phone art without failing Apple's pixel check.

Marketing render versus raw UI: the honest line

There are two schools. Raw UI means you upload the plain screen straight from the simulator. Marketing render means you place that screen inside a device frame on a designed background with a caption above it. Marketing renders convert better for almost every category, and they are allowed, with one hard rule: the screen inside the frame has to be a real screen from your app.

We built a dedicated Screenshot Mode into SparkQuest for exactly this reason. It puts the app into a clean, repeatable state so every capture is a genuine screen rather than a mockup drawn in a design tool. That is the honest version of a marketing render: real pixels, framed nicely. The line you cannot cross is a panel that is all background art and headline with no actual interface, or a screen showing a feature you have not shipped. Both are rejectable in 2026, and both are easy for a reviewer to catch.

Caption craft: three to seven words, and now they are indexed

Keep each caption to three to seven words, one idea per panel, legible at thumbnail size. If a caption needs two reads, it is too long. This was always a design rule for humans, but since June 2025 it is also a ranking rule: Apple runs OCR across your screenshots and indexes the text it finds, so your captions feed keyword relevance the same way your subtitle does. Write them to be true, clear, and keyword-honest, and they work twice.

Here is the honest lesson from our own listing. Our first caption set for SparkQuest read "Generate AI App Ideas in Seconds," "Hunt for LEGENDARY Concepts," "Copy-Paste Into Any AI Builder," "Calculate Your Path to $10K/Month," and "Build Your App Empire." The first three are fine: descriptive, real, tied to actual screens. The fourth is a review risk. Dollar-figure earnings overlays ("$10K/Month") are one of the most common creative rejections in 2026, right alongside a listing that carried a fabricated "Join 10,000+ builders" line before we removed it in an honesty sweep. Descriptive captions that name what the screen does clear review and age better than a claim you would have to defend. We changed our approach for that reason, not because a reviewer forced us to.

The review traps that kill creative

The four creative rejections that actually happen in 2026, in rough order of frequency:

• Wrong dimensions. One pixel off the required size. Purely mechanical, purely avoidable.
• Pricing or earnings text baked into the image. "$10K/month," "50% off," a subscription price rendered as art. Pricing belongs in the price field, not the screenshot.
• Comparative claims naming a competitor. "Better than [X]" in a panel gets flagged.
• Misleading content. A screen showing a feature, a paywall, or a state that is not in the shipped build.

None of these are about taste. They are about the reviewer being able to match every claim to something real in the binary. Keep the captions descriptive and the screens genuine and you sidestep all four. If a rejection does land, our guide to App Store rejection reasons in 2026 walks the exact letters and fixes.

Per-store requirements, side by side

The two stores want different things, and a set built for one needs adjusting for the other.

• Orientation. Portrait wins for most apps on both stores, because portrait panels show more of themselves in search results and mirror how people hold a phone. Landscape earns its place only in gaming and streaming, where immersion is the pitch.
• Count. Up to 10 panels per device family on iOS, up to 8 per type on Google Play. Neither is a target; three strong panels beat ten thin ones.
• Format. iOS wants PNG or JPEG, RGB, no alpha channel. Google Play accepts JPEG or 24-bit PNG in a 320-to-3840-pixel range with a 2:1 or 16:9 aspect band.
• Caption indexing. iOS runs OCR on your captions and indexes the text (since June 2025). Google Play does not index screenshot text, so on Play the caption is purely a human pitch.
• Feature graphic. Google Play additionally wants a 1024 by 500 feature graphic at the top of the listing, which iOS has no equivalent for. We built one for SparkQuest; it is a separate asset from the screenshots, not a resized panel.

The practical workflow: design the set at the iPhone 6.9-inch spec first, because it is the tightest and most consequential, then re-export for the 13-inch iPad at 2064 by 2752 and letterbox a landscape or square variant for Play. The words on the panels can stay the same; only the frame and the export size change.

Order, localization, and testing the set

Three things separate a competent set from one that actually lifts installs.

Order is a lever, not a given. The single most testable decision is which panel goes first. Two sets with identical art in a different order convert differently, so if you can only test one thing, test panel one. Lead with the screen that states the value most plainly, not the one that looks prettiest.

Localization is indexed real estate on iOS. Because iOS reads your captions with OCR and indexes them, translated captions in each localized storefront feed keyword relevance in that language, not just readability. If you localize, translate the captions properly rather than leaving English art in a French store, because a mistranslated or clipped caption both reads badly and indexes badly. This is a real 2026 caption trap: rushed machine translation that clips a headline mid-phrase.

Test where you can measure. iOS Product Page Optimization and Google Play store listing experiments both let you A/B test screenshot sets against live traffic and read a real conversion difference. Ship a hypothesis, hold it for a full measurement window, and let the store tell you which set wins instead of guessing in a design tool.

Tools that get you there

You do not need a design agency. The native iOS Simulator captures pixel-correct screens with a keyboard shortcut, which handles the raw layer. For the framed marketing render, a purpose-built Screenshot Mode in your own app (like ours) guarantees real states, and any layout tool can drop that capture into a device frame with a caption. Dedicated screenshot builders exist if you want templates, but the honest core of the job is the same either way: real screens, clear captions, correct sizes. The whole set for both stores is a half-day of work once your states are clean, and it is worth budgeting that half-day, because this is one of the highest-leverage surfaces in the entire launch.

Polish comes after the idea is real

A great screenshot set cannot rescue an app nobody wants, and it is the last mile, not the first. Before you spend a day on renders, be honest that the idea earned the build: run it through the same demand test our forge applies when it grades an idea from COMMON to LEGENDARY, and read how to validate an app idea so the panels you are polishing are selling something people asked for.

When the app is real and the states are clean, screenshots are where a lot of quiet installs are won or lost. Get the first three telling one honest story, size them to the pixel, keep the captions descriptive, and you have done the part most builders skip.

This guide is one stop on the full path. For the whole sequence from finished build to live on both stores, start at how to ship an app in 2026, and pair this with the exact App Store screenshot sizes and specs and the ASO guide that explains how your indexed caption text feeds ranking.