Translation and localisation industry statistics, 2026
Market size, how buyers react to language, app store reach, and what translation platforms charge. Every figure links to its source, and the app localisation numbers come from our own measurement of 338,859 shipped UI translations in 27 languages.
Updated . Each statistic has its own link, so you can cite it directly.
Ten numbers worth knowing
- The global language industry was worth an estimated USD 72.6 billion in 2025, projected to reach USD 73.4 billion in 2026.
- The 100 largest language service and technology providers made up only 19.8% of the language industry in 2025, and the top 10 just 9.9%.
- 76% of online shoppers prefer to buy products with information in their native language.
- 40% of online shoppers will never buy from websites in other languages.
- 49.5% of websites whose content language can be identified use English; the next largest are Spanish (6.1%), German (5.9%), Japanese (4.9%) and French (4.6%).
- Across 14 European languages, UI text runs a median 18% longer than English, from +9% (Danish and Norwegian) to +32% (French).
- 1 in 13 French UI strings and 1 in 17 German strings are at least twice as long as the English, so a layout sized exactly for English will truncate or wrap them.
- Chinese (Simplified) UI strings use 66% fewer characters than English, and Japanese and Korean 45–46% fewer, although each character takes more horizontal space on screen.
- About 1 in 950 shipped translations that contain a placeholder drop, add or break it (47 of 44,495), across three mature, widely translated open-source apps.
- 4 of the 5 competitor TMS offers we compared (XTM, Phrase, Lokalise and Smartling) meter AI translation or AI quality checks separately from processed words; only Smartcat uses one unit, its Smartwords.
Market size and growth
The two most-cited market sizes measure different things. Nimdzi counts the whole language industry, including interpreting and services. Slator counts the addressable market for language solutions and AI that buyers spend on.
$72.6bn
Language industry size, 2025
The global language industry was worth an estimated USD 72.6 billion in 2025, projected to reach USD 73.4 billion in 2026.
<1%
Projected annual growth to 2030
Nimdzi projects language-industry growth of under 1% a year from 2026, reaching USD 76.1 billion by 2030 in its realistic model and USD 96.9 billion in its optimistic one.
19.8%
Share held by the 100 largest providers
The 100 largest language service and technology providers made up only 19.8% of the language industry in 2025, and the top 10 just 9.9%.
+1.3%
Combined growth of the top 100, 2024–2025
The Nimdzi 100 grew by a combined 1.3% from 2024 to 2025; 53 of the 100 providers grew and 22 shrank.
$30.85bn
Language solutions and AI market, 2025
Slator values the addressable market for language solutions and AI at USD 30.85 billion in 2025, growing to USD 36.10 billion by 2031 (2.65% CAGR).
Slator, 2026 Language Solutions & AI Market Report, 2026
Slator’s figure covers buyer spend on translation, localisation, dubbing, subtitling, captions, multilingual content generation, accessibility, speech translation and real-time multilingual communication.
Buyers, consumers and the web
Why localisation pays: what consumers say about language, and how unevenly content is spread across languages.
76%
Online shoppers who prefer their own language
76% of online shoppers prefer to buy products with information in their native language.
CSA Research, Can't Read, Won't Buy – B2C (8,709 consumers, 29 countries), 7 July 2020
40%
Never buy from sites in other languages
40% of online shoppers will never buy from websites in other languages.
CSA Research, Can't Read, Won't Buy – B2C (8,709 consumers, 29 countries), 7 July 2020
65%
Prefer their language, even if quality is poor
65% of consumers prefer content in their own language even if it is of poor quality.
CSA Research, Can't Read, Won't Buy – B2C (8,709 consumers, 29 countries), 7 July 2020
75%
More likely to buy again with support in their language
75% of consumers are more likely to buy from the same brand again if customer care is in their language.
CSA Research, Can't Read, Won't Buy – B2C (8,709 consumers, 29 countries), 7 July 2020
49.5%
Websites with English content
49.5% of websites whose content language can be identified use English; the next largest are Spanish (6.1%), German (5.9%), Japanese (4.9%) and French (4.6%).
W3Techs, Usage statistics of content languages for websites, 23 October 2026
App and UI localisation
Store reach from Apple, plus our own measurement of shipped UI translations in 27 languages.
How much longer translated UI text runs than English
Change in visible characters, 338,859 shipped translations in Mastodon, Signal and WordPress. The thin line spans the lowest to highest of the three apps.
- French +32%
- Greek +30%
- German +28%
- Spanish +24%
- Italian +23%
- Dutch +20%
- Hungarian +20%
- Russian +17%
- Portuguese (Brazil) +16%
- Polish +16%
- Ukrainian +14%
- Finnish +12%
- Hindi +12%
- Swedish +11%
- Czech +11%
- Indonesian +11%
- Danish +9%
- Norwegian +9%
- Turkish +9%
- Vietnamese +7%
- Thai −1%
- Arabic −4%
- Hebrew −17%
- Japanese −45%
- Korean −46%
- Chinese (Traditional) −62%
- Chinese (Simplified) −66%
175
App Store countries and regions
The App Store serves over 850 million average weekly users across 175 countries and regions.
Apple Newsroom, study by Analysis Group, 4 June 2026
50
App Store listing languages
App Store Connect supports localised app metadata in 50 languages, after adding 11 in March 2026, ten of them Indian languages.
Apple Developer News, 31 March 2026
+28%
German UI text vs English
German UI strings are 28% longer than the English source on average (27–29% across Mastodon, Signal and WordPress; 13,135 shipped translations).
Original research LanguageOps UI string study, 23 October 2026
+32%
French UI text vs English
French UI strings are 32% longer than English, the most of the 27 languages measured; 28.3% of French strings grow by more than half.
Original research LanguageOps UI string study, 23 October 2026
+18%
Median expansion, 14 European languages
Across 14 European languages, UI text runs a median 18% longer than English, from +9% (Danish and Norwegian) to +32% (French).
Original research LanguageOps UI string study, 23 October 2026
1 in 13
French UI strings at least twice as long
1 in 13 French UI strings and 1 in 17 German strings are at least twice as long as the English, so a layout sized exactly for English will truncate or wrap them.
Original research LanguageOps UI string study, 23 October 2026
−66%
Chinese (Simplified) characters vs English
Chinese (Simplified) UI strings use 66% fewer characters than English, and Japanese and Korean 45–46% fewer, although each character takes more horizontal space on screen.
Original research LanguageOps UI string study, 23 October 2026
53–65%
UI strings of 20 characters or fewer
53–65% of app UI strings are 20 characters or fewer, with a median of 16–19 characters, so most strings give translators very little text to work with.
Original research LanguageOps UI string study, 23 October 2026
1 in 8
UI strings containing a variable
11.8–14.2% of app UI strings contain at least one variable placeholder such as %s or {name}.
Original research LanguageOps UI string study, 23 October 2026
1 in 950
Shipped translations with a broken placeholder
About 1 in 950 shipped translations that contain a placeholder drop, add or break it (47 of 44,495), across three mature, widely translated open-source apps.
Original research LanguageOps UI string study, 23 October 2026
338,859
Shipped translations measured
The LanguageOps UI string study measured 338,859 shipped translations of 13,682 English UI strings in 27 languages across Mastodon, Signal and WordPress.
Original research LanguageOps UI string study, 23 October 2026
TMS pricing and AI allowances
Our analysis of published plans and supplied quotes for a production translation workflow.
4 of 5
Competitor TMS plans with a separate AI meter
4 of the 5 competitor TMS offers we compared (XTM, Phrase, Lokalise and Smartling) meter AI translation or AI quality checks separately from processed words; only Smartcat uses one unit, its Smartwords.
Our analysis LanguageOps TMS pricing comparison, 15 July 2026
Based on published plans and supplied quotes for XTM Cloud, Phrase, Lokalise, Smartling and Smartcat, reviewed 15 July 2026.
15%
AI words vs processed words, Lokalise Advanced
Lokalise Advanced includes 150,000 Pro AI words a year alongside 1 million processed words, so its Pro AI allowance covers 15% of the words it lets you process, or 7.5% of a pre-translate-plus-LQA workload.
Our analysis LanguageOps TMS pricing comparison, 15 July 2026
150,000 ÷ 1,000,000 = 15%. A two-pass workload (AI pre-translation, then AI LQA) needs 2 million AI words.
$12k–$60k+
Competitor list price, 1M-word workflow
For a 1-million-word annual workflow, competitor list prices ran from $11,988 a year (Lokalise Advanced) to over $60,000 (Smartling, whose AI translation alone starts at $0.06 a word), against £3,000 for LanguageOps Professional.
Our analysis LanguageOps TMS pricing comparison, 15 July 2026
List prices in each vendor’s own currency, not converted. Smartling’s figure is 1M words at its published $0.06 starting AI rate, before its unlisted TMS price.
How much longer translated UI text runs: 27 languages, three apps
We compared the English source with every shipped translation in Mastodon’s web app, Signal for Android and WordPress core: 338,859 translations of 13,682 English strings. The three apps use different formats and translation communities, yet agree within six percentage points for 17 of the 27 languages.
| Language | All strings vs English | Mastodon | Signal | WordPress | Strings >50% longer | Strings 2× or longer | Short strings (≤10 chars), median | Translations |
|---|---|---|---|---|---|---|---|---|
| French fr | +32% | +29% | +35% | +31% | 28.3% | 7.5% | +40% | 13,132 |
| Greek el | +30% | +31% | +33% | +29% | 27.5% | 7.7% | +44% | 13,072 |
| German de | +28% | +29% | +29% | +27% | 22.1% | 6% | +33% | 13,135 |
| Spanish es | +24% | +21% | +26% | +24% | 22% | 5.6% | +30% | 13,262 |
| Italian it | +23% | +22% | +26% | +22% | 18.2% | 4.5% | +25% | 13,089 |
| Dutch nl | +20% | +22% | +27% | +17% | 16.2% | 4.1% | +33% | 12,962 |
| Hungarian hu | +20% | +19% | +25% | +18% | 20.2% | 6.2% | +29% | 12,558 |
| Russian ru | +17% | +23% | +23% | +14% | 18% | 6.3% | +30% | 13,355 |
| Portuguese (Brazil) pt-BR | +16% | +12% | +21% | +14% | 14.7% | 3.4% | +25% | 12,723 |
| Polish pl | +16% | +12% | +24% | +12% | 16.7% | 5.4% | +29% | 13,018 |
| Ukrainian uk | +14% | +13% | +18% | +12% | 18.5% | 6.4% | +33% | 11,690 |
| Finnish fi | +12% | +14% | +13% | +11% | 14.8% | 4.8% | +25% | 13,327 |
| Hindi hi | +12% | +9% | +17% | +8% | 12.8% | 4.5% | +11% | 10,236 |
| Swedish sv | +11% | +8% | +13% | +10% | 8.9% | 2.4% | +12% | 13,217 |
| Czech cs | +11% | +9% | +11% | +11% | 12.8% | 4.2% | +25% | 13,163 |
| Indonesian id | +11% | +13% | +14% | +8% | 10.2% | 3.2% | +20% | 11,217 |
| Danish da | +9% | +7% | +12% | +8% | 6.9% | 1.9% | +11% | 12,585 |
| Norwegian nb | +9% | +7% | +12% | +8% | 8% | 2.1% | +11% | 12,588 |
| Turkish tr | +9% | +16% | +9% | +9% | 11.8% | 3.8% | +20% | 13,318 |
| Vietnamese vi | +7% | -2% | +11% | +7% | 11.9% | 3.5% | +17% | 12,663 |
| Thai th | -1% | -2% | +4% | -5% | 9.2% | 2.8% | +17% | 9,315 |
| Arabic ar | -4% | -5% | +3% | -7% | 8% | 2.8% | +10% | 13,139 |
| Hebrew he | -17% | -17% | -14% | -19% | 2.7% | 0.8% | 0% | 10,639 |
| Japanese ja | -45% | -46% | -44% | -45% | 0.2% | 0.1% | -40% | 12,827 |
| Korean ko | -46% | -47% | -44% | -46% | 0.3% | 0.1% | -50% | 12,846 |
| Chinese (Traditional) zh-TW | -62% | -63% | -64% | -60% | 0.5% | 0.3% | -56% | 12,383 |
| Chinese (Simplified) zh-CN | -66% | -65% | -65% | -66% | 0.1% | 0.1% | -60% | 13,400 |
Sorted by expansion. “All strings” compares total characters; the short-string column is the median ratio for English strings of ten characters or fewer. Download as CSV.
What we measured
| App | Format | English strings | With a variable | ≤20 characters |
|---|---|---|---|---|
| Mastodon (web app)131aabcaf1f1, 23 October 2026 | JSON, ICU messages | 1,520 | 14% | 64.6% |
| Signal (Android)16f138410a4a, 22 October 2026 | Android strings.xml | 4,139 | 11.8% | 58.8% |
| WordPress (core)WordPress - 7.1.x - Development, 23 October 2026 | gettext PO | 8,023 | 14.2% | 52.7% |
Placeholder defects that shipped
47 of 44,495 translations that contain a placeholder drop, add or misnumber it, about 1 in 950. A further 19 change only the type, such as %d written as %s, and are not counted. Some drops are deliberate, where a translator typed the value in. Others lose information or print a literal code:
- cs · WordPress
%1$d of %2$d active%1 z %2 aktivních“%1” and “%2” are not valid placeholders, so the counts are lost. - cs · WordPress
+%dprocentA count such as “+5” became the word “percent”. - es · Signal
%1$s raised a handHas levantado la manoReads “You have raised your hand”: the person's name is gone. - ru · Mastodon
Created by {name}Созданные пользователемReads “Created by the user”: which user is no longer shown.
How the study works
We took the English source strings and every shipped translation from three widely translated open-source apps, each in a different file format: Mastodon’s web app (JSON, ICU messages), Signal for Android (Android strings.xml) and WordPress core (gettext PO from translate.wordpress.org). For 27 languages, we compared the visible length of each translation with its English source, after removing placeholders and markup, and checked whether the placeholders survived.
Translations identical to the English, empty or marked fuzzy were left out, as were plural and select messages, whose forms legitimately differ. Expansion is the total translated characters divided by the total English characters. Placeholders are compared by position, so a legal reorder such as %s … %s → %2$s … %1$s is not an error.
Character counts are not the same as screen width. Chinese, Japanese and Korean use fewer characters than English, but each one takes more horizontal space, so measure rendered width in your own UI before shrinking a layout. The measurement script is in our site repository and can be re-run against newer versions of each app.
What it means for app teams
- Leave room to grow. French, Greek and German UI text runs 28–32% longer than English, and individual short labels often double.
- Give translators context. More than half of UI strings are 20 characters or fewer, too short to translate reliably without the key name, a description or a screenshot.
- Check placeholders before you ship. About 1 in 8 UI strings contains a variable. Even in mature projects, about 1 in 950 shipped translations drops or breaks one.
See how LanguageOps handles this on the app localization page, or compare platform costs in the TMS pricing comparison.
Exact rules used by the measurement script
- Singular strings only. ICU plural/select messages, Android <plurals>/<string-array> and
gettext msgid_plural entries are skipped, because their forms legitimately differ.
- translatable="false", fuzzy, empty and target-identical-to-source strings are excluded.
- Visible length = characters left after removing placeholders and markup tags, with
whitespace collapsed. Ratios are measured only where the source has >= 2 visible characters.
- Placeholder integrity: the target must carry the same placeholders as the source. printf
placeholders are normalised to (position, conversion), so a legal reorder
(%s %s -> %2$s %1$s) is not an error. A %d written as %s counts as "type only", not as a
defect. ICU compares the set of {argument} names.
Where every number comes from
Third-party figures were checked against the publisher’s own page on the date shown. Our own analysis and research is marked as such on each card.
- Nimdzi Insights, The 2026 Nimdzi 100, 2026. Checked 23 October 2026.
- Slator, 2026 Language Solutions & AI Market Report, 2026. Checked 23 October 2026.
- CSA Research, Can't Read, Won't Buy – B2C (8,709 consumers, 29 countries), 7 July 2020. Checked 23 October 2026.
- W3Techs, Usage statistics of content languages for websites, 23 October 2026. Checked 23 October 2026.
- Apple Newsroom, study by Analysis Group, 4 June 2026. Checked 23 October 2026.
- Apple Developer News, 31 March 2026. Checked 23 October 2026.
- LanguageOps TMS pricing comparison, 15 July 2026.
- LanguageOps UI string study, 23 October 2026.
How to cite this page
Our original research and analysis are published under CC BY 4.0: use them freely with a link back. For third-party figures, cite the original publisher, which each card names. Every card has a Copy citation link and its own anchor.
LanguageOps (2026). UI string expansion and placeholder integrity in shipped app translations. https://languageops.com/translation-industry-statistics/#ui-study
Machine-readable: https://languageops.com/research/translation-industry-statistics.json · https://languageops.com/research/ui-string-expansion-2026.csv
Translation statistics, answered
How big is the translation industry?
How much longer is German than English?
Which language expands the most when translated from English?
How much space should I leave for translated UI text?
What percentage of consumers prefer to buy in their own language?
What percentage of the web is in English?
Can I use these statistics?
Measure your own strings
Send a resource file and your target languages. We will show you how much your text grows, where placeholders are at risk, and what the translated files look like.