Localizing keywords vs translating your App Store listing: why they're not the same

Type "workout" into Google Translate and ask for Japanese. You get いい結果 ("a good result"). Ask for Brazilian Portuguese and you get dar certo ("to turn out fine"). We got both on September 27, 2026, and again the next morning. That's the funny failure, and it's rare. The costly one is quieter: 予算アプリ is a correct Japanese translation of "budget app," yet 22 of the 23 top Japanese results for the category call themselves 家計簿 ("household account book").

That gap is why localizing keywords and translating your App Store listing are different jobs. Search is where most installs start. Apple says almost 65% of downloads happen directly after a search (Apple Ads, citing 2022 worldwide data). Each localization then ranks on three short fields: a 30-character name, a 30-character subtitle and a 100-character keyword field (Apple's product page guide). Translate those fields and you fill them with words that are correct, and often not the ones people type.

Below: how often translation missed in our 84-check test, what locals type instead, a research workflow that works without speaking the language, and the fields where translation is fine. For which languages to localize first and what localization costs, see our separate guides. If you'd rather hand the research off, it's the core of our app localization service.

What's the difference between translating a listing and localizing keywords?

Translation converts your English meaning into another language. Keyword localization starts at the other end. It finds what people in one storefront type when they want an app like yours, then builds the name, subtitle and keyword field around those words. Apple draws the same line, telling developers to "tailor your keywords to reflect the values of each market" (Apple Developer).

The two methods answer different questions. Translation asks what your copy means in German. Localization asks what Germans type when they need what your app does. Sometimes the answers match. Often they don't. When that happens, a fluent, accurate keyword field still misses the query that matters.

The stakes are higher in ASO than on the web. The space is tiny. Each App Store localization gets 160 indexed characters across name, subtitle and keywords, and the description is widely reported not to count for App Store ranking (see which fields each store indexes). So every slot has to hold a real query.

Apple's reach multiplies the job. "The App Store is available in 175 regions and 50 languages" (Apple Developer, Localization), and each storefront reads only the localizations Apple assigns to it. So the unit of work is the storefront, not the language. Translating a whole listing is the cheap part. The research is where the hours go.

How often does a straight translation miss what locals search?

About half the time. In our September 27, 2026 check, we ran 12 common app keywords through Google Translate for 7 App Store storefronts. Apple's own autocomplete (the App Store search suggestions) returned the translation verbatim in 44 of 84 checks (52%), and it was the first suggestion in only 23 (27%). For 10 outputs (12%), it suggested nothing at all.

Then we classified each output, using autocomplete plus the titles of the apps that rank. By our judgment, which native speakers haven't yet reviewed, 50 of the 84 were fine. Another 31 (37%) were correct translations that locals don't mainly type. Just 3 (4%) were simply wrong: the "workout" pair, いい結果 and dar certo, and Brazil's cartões for "flashcards," where autocomplete returns bank and store cards.

Some patterns held across markets. "Budget app" wasn't the local term in any of the seven storefronts (0 of 7): Japan says 家計簿, Germany Haushaltsbuch, and Spain and Mexico control de gastos. At the other end, meditation, step counter and VPN were fine everywhere (21 of 21). Some words really do just translate. The last section covers which.

By storefront, Brazil's translations did best (9 of 12 fine) and French (Canada) worst (5 of 12). One reason French did worst in a French-speaking market: Canada's search box isn't only French (see the regional section).

What do German and Japanese users type instead of the translation?

They type the category's native name. That's often a compound or a kanji term, and sometimes plain English. In Japan's top results for 家計簿, 22 of 23 app titles use 家計簿. Only 2 use the translation's 予算 ("budget"). In Germany, 19 of 24 top titles for Haushaltsbuch use Haushalts- terms (our September 27, 2026 check).

Even those two 予算 titles put the word next to 家計簿, and the table shows the same gap across the German and Japanese keywords we tested.

German: compounds, and English that's already local

German builds category names from compounds. The budget word is Haushaltsbuch, and photo editing is Bildbearbeitung or Foto bearbeiten. The translation Bildbearbeiter means an image editor as a person. Its suggestions were mostly app names. Where autocomplete shows a compound and a split form, like Fotobearbeitung and Foto bearbeiten, research both.

English is a native query too. German autocomplete returned "habit tracker," "to do list," "workout tracker" and "dating app" as queries. A German-only keyword field misses those. Germany also lists the English (U.K.) localization, so the loanwords can live there and leave the German field for compounds (more on shared fields below).

Length is the other trap. IBM's guideline, cited by the W3C, puts European-language translations of a 21–30-character English string at 160–180% of its length (W3C Internationalization, Text size in translation). By our arithmetic, a full 30-character English name would need 48–54 characters in German. The limit is 30. So German names get rewritten around the head term, not translated.

One field also serves four storefronts. Apple lists German in Germany, Austria, Switzerland and Luxembourg. Swiss autocomplete suggests velowege schweiz, using the Swiss Velo rather than the German Fahrrad for bike.

Japanese: native category names, and more than one script

Japan's category names often have no English twin. The table shows how rarely the translation reaches app titles. Two details don't fit in it, though. First, dating apps also sit under 恋活 and 婚活 (dating and marriage hunting), and the translation 出会い系アプリ is an older term with a different connotation. Second, habit and to-do apps lean on 習慣化 ("habit-forming") and タスク管理 ("task management").

Script matters as well. People type かけいぼ, the hiragana reading of 家計簿. It autocompletes on its own, with variants like かけいぼ 無料 ("free"). Eight of the 23 top 家計簿 titles add that reading. Density works in your favor, though. The word 家計簿 takes 3 characters where "budget app" takes 10, so a 100-character Japanese keyword field holds far more concepts. Fill it with concepts, not padding.

Why don't Spanish, Portuguese and French keywords transfer between regions?

Because Apple treats each variant as its own localization, with very different reach, and translation tools mostly don't. Apple's table lists Spanish (Mexico) in 19 storefronts, including the US, and Spanish (Spain) in one. Yet Google Translate returned the same Spanish for es, es-MX and es-419 on all 12 phrases we tested (September 27 and 28, 2026).

Spanish: one field for 19 storefronts, one for Spain

Spain and Mexico split on everyday words. Grocery lists are lista de la compra in Spain and lista de super or lista del super in Mexico. Used cars are coches segunda mano in Spain and seminuevos in Mexico, while the translation autos usados returned a single app name as its only suggestion in both. Spain's to-do apps are lista de tareas, and Mexico's autocomplete adds lista de pendientes.

Spanish (Mexico) isn't just Mexican Spanish. Apple lists it in Mexico, the US and 17 other storefronts across Latin America. For lista de, Colombia's autocomplete suggests lista de mercado and Chile's lista de supermercado, while Argentina's shows both lista de supermercado and lista del super. All four storefronts we checked, Mexico included, also suggest lista de compras, the machine translation.

In a field that serves 19 markets, that neutral term can be the right anchor, with the variant for your largest market beside it. Spain's field serves only Spain. It can be fully local. The US reads Spanish (Mexico) too, and our guide to which keyword fields each storefront indexes covers that overlap.

Portuguese: Brazil reaches two storefronts, Portugal one

Apple lists Portuguese (Brazil) in Brazil and the US, and Portuguese (Portugal) only in Portugal. Brazil's autocomplete says academia (gym), treino and controle financeiro. Portugal's says treino de ginásio and telemóvel. Google's European Portuguese differed from its Brazilian output on just 1 of 12 phrases, and it even kept aplicativo, the Brazilian word, where European Portuguese usually says aplicação.

French: 32 storefronts for France's field, one for Canada's

Apple lists French in 32 storefronts and French (Canada) only in Canada, beside English (Canada). Google's Canadian French is region-aware (voitures usagées for used cars), but its generic French gave France the Quebec term liste d'épicerie, where French autocomplete says liste de course(s).

Canada surprised us. In our check, its storefront returned the same suggestions whether we tagged the request French or English, and "budget" returned English phrases (for example, "budget planner" and "budget tracker"). So your French (Canada) field may compete in a search box that leans English, which makes Quebec-specific terms worth more than generic French.

The same storefront-first logic applies to India. Apple added 11 metadata languages on March 31, 2026, "especially in India" (Apple Developer News), and its table now lists 12 localizations for India's storefront. Our guide to App Store Connect's new Indian languages covers how to test them.

How do you research native keywords country by country?

Run the same seven steps in every storefront. Start from what your app does, not from your English keyword list. The evidence comes from the storefront: its autocomplete, top-ranking app titles, local reviews and one native reviewer. Apple Ads adds Search Popularity, shown "as numbers from 1 to 5, with 5 being the most popular" (Apple Ads glossary).

Here's how we run it. This covers only the cross-language layer. For scoring, difficulty and relevance, see the full keyword research workflow.

1. List concepts, not keywords. Write the 10–15 jobs your app does ("track spending," "share a shopping list") in plain English. Translators translate words. You want each market's word for the job.
2. Harvest autocomplete in that storefront. Switch the App Store to that country, or use a tool that queries it. Type the first few characters of each concept in every script and form locals use. That means German compounds and split forms, Japanese kanji and kana, and English loanwords everywhere. Log the suggestions and their order. They show popularity, not volume.
3. Count the words in the top apps' names and subtitles. Open the top 10 for each candidate and tally the terms they use, as the chart above does. If 18 of 23 titles use 買い物 ("shopping"), as Japan's shopping-list apps do, that's the category's word. Note brand names in autocomplete ("bring! liste de courses," "マッチングアプリ pairs"), but never copy them. Apple doesn't allow "competing app names" in keywords.
4. Size it with Apple Ads. Check candidates' Search Popularity for the campaign's country or region. Compare them with each other, not with English terms. If you have budget, a small Search Match campaign surfaces real local queries in the Apple Ads search terms report. Search Match draws on "the metadata from your App Store listing, information about similar apps in the same genre, and other search data" (Apple Ads Help).
5. Mine local reviews. Competitors' reviews in that storefront use the market's own vocabulary. That makes them a source of native keywords no translator will give you. The App Review Keyword Analyzer pulls each storefront's reviews separately.
6. Run the native reviewer pass. The next section has the checklist.
7. Place, pack and track. Put the head term in the name or subtitle and the rest in the keyword field (see our keyword field formatting rules). English loanwords go in the English field that storefront also reads. Check ranks with the Keyword Rank Checker at two and four weeks. Then repeat the research at your next metadata update.

You don't need a native speaker until step 6. Steps 1 to 5 run on public evidence, so the one person who speaks the language spends their time judging a shortlist, not translating a spreadsheet.

What should a native reviewer check before you ship?

A native reviewer should check six things that search data can't settle on its own: connotation, regional variant, loanword versus native term, script, brand names and plain wrong sense. That's a different job from proofreading a translation. Apple warns that "improper use of keywords is a common reason for App Store rejections" (Apple Developer, 2026), so the brand check isn't optional.

1. Connotation. Is it the word the category's leaders use? 出会い系アプリ versus マッチングアプリ, the term Japanese autocomplete pairs with leading dating brands such as Pairs and Tapple.
2. Region. Is it this storefront's word? Lista de compras versus lista de la compra in Spain; liste d'épicerie is Quebec, not France.
3. Loanword or native. Would a local search the English? "Habit tracker" in Germany, "flashcards" in Mexico, Spain and Brazil.
4. Script and form. Kanji or kana; a German compound or its split form.
5. Brand and trademark. Strip competitor names picked up from autocomplete.
6. Sense. Does the word mean the thing? Cartões for flashcards doesn't.

Brief them with evidence, not a word list. Send the autocomplete log, the title counts and the popularity scores. Then ask for a verdict on each candidate. Asking a reviewer what they'd type gets you more than asking whether a translation is correct.

When is translation alone good enough?

Translation alone is enough when the word isn't doing search work, or when locals already search the translated word. In our September 2026 check, meditation, step counter and VPN translated cleanly in all 7 storefronts (21 of 21). Ordinary prose is safer ground too. Intento's 2025 evaluation covered 46 engines and LLMs, and its human evaluators "often couldn't distinguish AI-generated from human translations" (vendor-published, via Slator, October 30, 2025).

• Translate, with a native post-edit: the App Store description, promotional text, What's New and in-app strings. People read these after they've found you.
• Translate, then check: standard category nouns that autocomplete confirms, such as VPN, Schrittzähler, 歩数計 and contador de pasos.
• Translate for now: markets you're only probing. A translated listing, or a localized custom product page as a market test, is a cheap demand check before you fund research.
• Localize, always: the App Store name, subtitle and keyword field.
• Localize on Google Play, description included: Play has no keyword field and indexes the full description.

The long copy still counts once people land on your page. In a 2020 CSA Research survey of 8,709 consumers in 29 countries, 76% preferred product information in their own language, and 40% never bought from websites in other languages (CSA Research, July 7, 2020). So translate them well, with a native post-edit.

Google Play needs more care. Play indexes the full description. Play Console's free machine translation covers the title, short and full description, and Google says these translations are "not reviewed or approved by humans" (Play Console Help). Since I/O 2026, "Gemini models enable Play Console to pre-populate your listings across different languages for your review" (Android Developers Blog, May 19, 2026). Treat both as drafts. Then put researched terms in the title and short description.

For a team with ten markets and one budget, the split is simple. Localize the three short App Store fields everywhere you sell. Translate the long copy. Spend native-reviewer time on the storefronts that already bring impressions.

Data appendix: all 84 checks

Each cell shows Google Translate's output for that storefront and whether it appeared in App Store autocomplete, followed by our verdict and, in brackets, the term locals use where it differs. A ✓ means the exact translation appears in autocomplete. A ✗ means suggestions exist but not the exact term, and ∅ means no suggestions.

Verdict A is fine. B is a correct translation that isn't the term locals mainly type. C is the wrong sense. Verdicts are our classification, pending native review. Mexico and Spain share one Spanish output because Google returned the same text for every Spanish variant.

Across all storefronts: ✓ 44 of 84 (52%), first suggestion 23 (27%), no suggestions 10 (12%). Verdicts: A 50, B 31, C 3.

Title counts (September 27). Each row counts app titles that contain a term's core word. The titles come from the top iTunes Search API results for the local term, and each row compares the local word with Google Translate's word.

Both 予算 titles also contain 家計簿, and 8 of the 23 add the reading かけいぼ (re-checked September 28).

Re-checks and variants (September 28). Google Translate still gave いい結果 and dar certo for "workout" alone. With context, "workout app" gave トレーニングアプリ and aplicativo de treino. Suggestions for lista de included lista de super and lista del super in Mexico. Colombia showed lista de mercado, and Chile lista de supermercado. Argentina showed both lista de supermercado and lista del super. All four also showed lista de compras. Google's Spanish was the same for es, es-MX and es-419 on 12 of 12 phrases.

On September 27, Google's European Portuguese differed from its Brazilian output on 1 of 12 phrases, and its Canadian French differed from its generic French on 2 of 12.

To cite this data: ASO Agency, "Localizing Keywords vs Translating Your App Store Listing: Why They're Not the Same," asoagency.io, September 28, 2026.

The bottom line

Translation and keyword localization do different jobs. The App Store's three short fields only reward one of them.

• The difference: translation tells people what your app does. Localization matches what they type.
• The miss rate: 44 of 84 translations matched autocomplete. Wrong words are rare; correct words locals don't mainly type are common (our classification).
• The variants: Spanish (Mexico) covers 19 storefronts and Spanish (Spain) one, so write each field for everyone who reads it.
• The method: seven steps per storefront, with a native reviewer judging evidence rather than translating.
• The split: translate descriptions and release notes; localize the name, subtitle, keyword field and everything Google Play indexes.

See which keyword fields your storefronts read in the free Cross-Localization Explorer. Then book a free 30-minute call and we'll run the per-country research for your first markets, or start with the fixed-price ASO audit.