App Store Screenshot Localization — Ship 10 Languages Without 10x the Work

Localised screenshots are one of the highest-return changes available in app marketing, and one of the least done, for an obvious reason: doing it the naive way means rebuilding your entire set once per language, forever, every time anything changes.

This guide covers which languages to pick, what actually needs translating, the text expansion problem that breaks otherwise good designs, and how to structure the work so that adding a sixth language costs a fraction of the first.

Why Localised Screenshots Move the Needle

Store browsing is a fast, low-attention activity. A user scanning search results gives your listing a couple of seconds, and in that window the images do nearly all the work. When the caption is in a language they have to mentally translate, you have spent their attention on comprehension rather than persuasion.

Three structural reasons localised images outperform localised text alone:

• Screenshots are visible in search results. On the App Store, the first images appear directly in the results list, before anyone taps through. The description only exists after a tap, and after a scroll.
• Images carry the promise. Your captions state the benefit. That is the part that needs to land in the user's language, more than your feature paragraph does.
• It signals commitment. A listing in the local language reads as a product that supports that market, which affects trust in exactly the categories where trust matters most, such as finance and health.

The effect is largest where English proficiency is lowest. Japan, South Korea, Brazil, Germany, France, Italy, Spain and Latin America generally reward localisation heavily. Nordic markets, the Netherlands and Singapore tolerate English listings far better, so they are usually not where you start.

How Each Store Handles Localisation

The two stores treat this differently, and the difference shapes your workflow.

App Store. You add a localisation per language in App Store Connect, and each one carries its own name, subtitle, keywords, description, screenshots and preview videos. Screenshots are per localisation and per device class, which is the combination that multiplies fast. If a localisation has no screenshots of its own, users in that locale see a fallback rather than something tailored.

Google Play. Similar structure: per-language store listings with their own graphics, plus Play's automatic translation option for text, which you should not rely on for marketing copy. Play also has a separate feature graphic per language, which people routinely forget to localise.

A practical consequence on both stores: the language your listing is localised into affects which store search terms you can rank for in that market. Localised screenshots and localised keywords work together, so treat them as one project rather than two. Our ASO guide covers the keyword side.

Choosing Your First Five Languages

Do not start from a list of the world's largest app markets. Start from your own data.

1. Open your store analytics and sort impressions by territory. You are looking for markets already finding you despite an English-only listing. That is demand you are currently converting badly.
2. Compare conversion rate by territory. A market with high impressions and a conversion rate well below your average is the clearest possible signal that language is the blocker.
3. Check competitive density. A localised listing in a market where competitors have not localised is a much larger edge than one more English listing at home.
4. Weigh monetisation, not just volume. Some markets deliver large install counts and very little revenue. If you sell subscriptions, weight by expected revenue per install.
5. Consider support cost. Localising the listing invites support requests in that language. Have at least a plan, even if the plan is machine translation with a clear disclosure.

For most apps this converges on a starting set of Japanese, German, French, Spanish and Brazilian Portuguese, with Korean and Simplified Chinese close behind. But confirm against your own numbers before committing.

Three Levels of Localisation

These are meaningfully different amounts of work, and the middle one is where most of the return sits.

Level 1 is the minimum viable version and worth doing immediately. Level 2 is where you should aim: an English UI behind a Japanese caption is a jarring contradiction that undercuts the caption's promise, so capture your screens with the app itself set to the target language. Level 3, replacing photography, names, currencies and cultural references, matters most in Japan, Korea and China, where localised-but-obviously-foreign listings underperform noticeably.

The Text Expansion Problem

This is what breaks designs. Translated text is rarely the same length as your English, and the direction is usually longer.

Two counterintuitive points. First, Japanese and Chinese being shorter does not mean you can shrink the type: complex glyphs need more pixels per character to stay legible, so keep the point size or raise it. Second, right-to-left languages do not just reverse the text, they reverse the reading order of the composition, so a device on the left with text on the right should usually mirror.

Designing a Set That Survives Translation

The cheapest localisation is a design that was built expecting it. Do these before you translate anything.

• Design captions to 60% of the available width. The leftover space absorbs German and French without a redesign.
• Allow two lines everywhere, even where English fits on one. Fix the block position, not the block height, so a second line grows downward without shifting the device.
• Never bake text into a background image. Keep captions as a separate editable layer, or you are re-rendering artwork per language.
• Choose a font with real language coverage. Many display fonts have no Cyrillic, Japanese or Arabic glyphs, and the silent fallback looks broken. Check coverage before you commit to a typeface.
• Avoid text inside device screens for the caption layer. Overlaying the interface means every translation needs recomposition.
• Keep captions to three to five words. Short lines expand less in absolute terms and translate more reliably.
• Avoid idioms and puns. They do not survive translation and translators will render them literally, which is worse than plain language.

A Workflow That Does Not Multiply

The goal is that language six costs a fraction of language one. That requires separating the parts that change from the parts that do not.

1. Lock the visual system once. Background treatment, device position, caption placement, type scale. This is identical across every language and should never be re-decided per locale.
2. Keep captions in a single source file. A spreadsheet or JSON file with one row per image and one column per language. This becomes the thing you hand to translators and the thing you diff when copy changes.
3. Capture localised UI in bulk. Set the simulator or device to each language and take the same screens in the same order. Name files consistently, for example home-de.png, so automation can find them.
4. Generate per locale rather than editing per locale. Whether that is a script, a Figma variable set, Fastlane, or a tool with localisation support, the rule is the same: never hand-edit the same layout twenty times.
5. Export with a naming convention that maps to store upload slots, so uploading forty images is a drag rather than a puzzle.
6. Review each locale visually before upload. Automation catches structure, not overflow. Look for clipped text, awkward line breaks and fallback fonts.

The engineering-heavy version of this is Fastlane with per-locale caption files, which is excellent if your team already lives in CI. The lighter version is a generator that accepts your caption file and produces each localised set from the same layout.

Getting the Translations Right

Marketing captions are the hardest text to machine translate well, because they are short, idiomatic and carry tone. A literal translation of a clever English line is usually flat or strange.

• Machine translation as a first draft is fine. Modern output is good enough to review, not good enough to ship unreviewed.
• Always have a native speaker check the captions. Five to ten short lines is a small, cheap job, and it is the single highest-value review in the entire localisation project.
• Give the translator context. Send the actual images, not a bare list of strings. "Track" translates differently for a fitness app than for a parcel delivery app.
• Specify a character budget per line. Translators will work to it if you tell them, and will not if you do not.
• Localise your keywords separately. Direct translations of your English keywords are rarely what people actually search in that market.
• Watch formality. German, Japanese and Korean have formality registers, and picking the wrong one makes an app sound either stiff or presumptuous. Tell the translator which you want.

Mistakes That Cost Conversions

• Translated captions over English UI. The most common half-done version, and it reads as untrustworthy.
• Shrinking type to fit a longer translation. Now the caption is both long and unreadable at thumbnail size. Cut words instead.
• Forgetting the Play feature graphic. Localised screenshots with an English feature graphic above them is a jarring mismatch.
• Localising the listing but not the first-run experience. You win the install and lose the user in the first thirty seconds.
• Leaving currency and units in US formats. Prices in dollars and distances in miles undercut everything else you localised.
• Never updating localised sets. Locales drift out of date because updating them is painful, so the markets you invested in end up with the stalest listings you have.

Measuring Whether It Worked

Localisation is measurable, so measure it rather than assuming.

• Compare conversion rate per territory before and after, not total installs, which move for many reasons.
• Roll out in waves. Localise two markets, wait for a few weeks of data, then decide on the rest. Untranslated markets act as a control group.
• Watch impressions too. Localised keywords should lift impressions in that territory, which is a separate effect from the conversion lift.
• Give it time. Store indexing and ranking changes take weeks, so a three-day read is noise.
• Check retention by locale. A conversion lift with poor retention means the listing is promising something the localised experience does not deliver.

For the broader conversion picture, see our guide to increasing App Store conversion rate.

Frequently Asked Questions