Localize in this order: Simplified Chinese, Japanese, Korean first; Brazilian Portuguese, Russian, Spanish, Turkish second; German, French, Traditional Chinese, Polish third. That ordering comes from two numbers per language — how many Steam users browse in it, and how few of them will buy an English-only game — and the rest of this post shows the math: current Steam language share, per-word market rates, postmortems with actual revenue impact (including one where localization flopped), and the $300 test you should run before spending $20,000.
One framing note up front. “Localization” hides two projects with a 50x cost difference: translating your store page (roughly 500-1,000 words, $100-300 per language) and translating your whole game ($3,000-5,000 per language for a mid-size indie at market rates). This post is about picking languages; our store page localization guide covers the Steamworks mechanics of shipping the cheap version first.
What the Steam language data says in 2026
The monthly Steam Hardware & Software Survey publishes a language breakdown. I pulled the June 2026 numbers on July 2:
Two provenance warnings before you budget off that chart. First, the survey samples whoever opts in during a given month, and it swings hard: the February 2026 survey briefly showed Simplified Chinese at 54.6%, a Chinese New Year sampling anomaly that corrected itself by March. Second, Valve’s own full-year 2024 account data, presented at GDC 2025, tells a stronger story for Chinese: Simplified Chinese passed English as the most-used language on Steam, 33.7% to 33.5%, with Russian third at 8.2%. However you slice it, the direction is the same — over 60% of Steam users run the client in a language other than English, a figure Valve states directly in the Steamworks localization documentation.
So an English-only game is invisible-ish to most of the platform. But raw share is only half the input, because plenty of non-English users buy English games anyway.
English tolerance: why share alone picks the wrong languages
The traditional localization set is FIGS: French, Italian, German, Spanish. On Steam that ordering is backwards, and Valve says so in its own docs: “the less a language looks like English, the higher the expected return.”
The best public evidence is a player survey the developers of Slay the Princess ran, covered in GameDiscoverCo’s localization data piece. They asked players in each region whether they would play the game in English if no localization existed:
Read those two charts together and the tier list writes itself. German is 2.7% of Steam and 61% of German players will buy your English build anyway, so German localization converts a minority of an already-small remainder. Simplified Chinese is 24-34% of Steam and only 32% of Chinese players tolerate English, so Chinese localization unlocks an audience most English-only games never touch. GameDiscoverCo’s estimate is blunt: without localization you can lose “90% or more” of potential players in some regions, China chief among them.
Japanese and Korean look tiny on the share chart (2.6% and 1.5%) but have the lowest English tolerance in the survey (23% and 22%) plus high spending power on premium games. Low tolerance times real spend beats raw share, which is how CJK ends up ahead of every European language.
The 2026 priority tier list
| Tier | Languages | Why |
|---|---|---|
| 1: do these first | Simplified Chinese, Japanese, Korean | Biggest gated audiences; English tolerance 22-32% |
| 2: strong seconds | Brazilian Portuguese, Russian, Spanish, Turkish | Large or fast-growing share; mid English proficiency; cheap rates |
| 3: cheap completers | German, French, Traditional Chinese, Polish | High English fluency, so smaller marginal gain, but low cost and low risk |
Notes on the edge cases:
- Russian is 8-10% of Steam and among the cheapest translations you can buy (see the rate table below). It sits in tier 2 only because English tolerance there is moderate and regional prices are low; on units it often performs like a tier-1 language.
- Spanish covers Spain plus Latin America, which browse Steam as separate language settings (4.59% and 0.81% in June 2026). One translation into neutral Spanish serves both reasonably well.
- Turkish is 1.32% of client share, low English proficiency, and translation runs at Russian-adjacent prices. A cheap option worth adding once tiers 1-2 are done.
- Traditional Chinese matters more if you plan a Switch or Asian console release; on Steam alone it is 1.5%.
If you can afford ten, GameDiscoverCo called Cobalt Core’s set “pretty much perfect” for an indie: English, French, German, Spanish, Japanese, Korean, Brazilian Portuguese, Russian, Simplified Chinese, Traditional Chinese.
What localization costs per word in 2026
Market rates for professional game translation cluster tighter than most devs expect. Allcorrect’s public pricing, a game-specialist vendor, quotes these per-source-word rates from English:
| Target language | Rate (EUR per word) |
|---|---|
| Russian | 0.06-0.09 |
| Simplified Chinese | 0.09-0.12 |
| Turkish | 0.09-0.12 |
| Korean | 0.09-0.14 |
| Brazilian Portuguese | 0.09-0.14 |
| Traditional Chinese | 0.09-0.14 |
| Japanese | 0.10-0.14 |
| German / French / Spanish | 0.10-0.14 |
Those are one vendor’s posted prices, not the whole market. Expect full-service agencies with editing and QA passes to quote above them, especially for CJK, and freelancers on direct contracts to come in a rough 30-50% under (my estimate from hiring, not a published rate card). Machine translation with native review is cheaper still, but shipping raw machine output for Chinese, Japanese, or Korean is a false economy — players spot it in minutes and review accordingly.
Now the budget math. Count your words honestly (UI strings, item descriptions, tutorials, dialogue) and multiply:
| Game size | Word count | Per language (at EUR 0.09-0.14) | Full 9-language set |
|---|---|---|---|
| Text-light (puzzle, action roguelike) | ~5,000 | ~EUR 450-700 | ~EUR 4,000-6,300 |
| Mid-size indie | ~30,000 | ~EUR 2,700-4,200 | ~EUR 24,000-38,000 |
| Text-heavy RPG / narrative game | 100,000+ | EUR 9,000-14,000+ | six figures |
That table is why the tier list matters. A text-light game can just buy all nine languages. A narrative game has to pick, and should pick by the tolerance data above, not by which languages feel familiar.
Game localization ROI: postmortems with numbers
The strongest aggregate data point is Chris Zukowski’s analysis of resting-week wishlist data collected from over 100 developers on howtomarketagame.com: during quiet non-promotional weeks, games with store pages localized beyond English earned a median of 62 wishlists per week versus 18 for English-only pages, about 3.4x. Zukowski flags the obvious confound himself — teams with translation budgets have bigger budgets for everything — so treat it as directional, not causal.
The individual postmortems are more concrete:
- Political Animals (Squeaky Wheel, 2016): after adding Simplified Chinese, units sold rose 3.2x across five post-launch weeks versus the five weeks prior (the dev excluded Winter Sale units from that comparison), and China was 17% of all sales in that Winter Sale, more than the UK and Canada combined. Chinese sales before localization were negligible. The dev’s caveat matters: revenue rose less than units because China’s recommended regional price is far below the US price.
- Against the Storm shipped with 17 supported languages, and China, Japan, Korea, and Taiwan together were at least 32% of sales as of late 2023, per GameDiscoverCo. The English-only contrast case in the same article, Nebulous: Fleet Command, sold over 125k units with China at just 2% and no other Asian country in its top 10.
- Florence (Mountains): lead designer Ken Wong reported that half its sales came from China, and more people played it in Mandarin than in English. Mostly mobile sales rather than Steam, but the language gate is the same.
And the honest counterexample: the developer of Wanba Warriors wrote up a localization push across dozens of store regions and concluded it “did not bring a noticeable increase in sales” — the December bump they saw tracked the holiday season, not the translations. Localization multiplies existing demand. A game nobody is finding in English does not become discoverable by existing in nine languages; fix the page and the wishlist pipeline first.
None of these figures are Valve-published. They are developer- and publisher-reported postmortems, which is the best data that exists in public. The consistent pattern across them — CJK unlocks the most, European languages the least — matches Valve’s own guidance.
Store-page-only localization: the $300 test
Before committing thousands per language to full localization, translate just the store page. Steam treats store localization and in-game localization as independent: you can publish a Simplified Chinese page for an English-only game, add a one-line disclaimer that in-game text is English for now, and watch what happens.
The mechanics, costs, and Steamworks setup are in our Steam page localization guide, but the test protocol is short:
- Translate the page copy (short description, long description, feature bullets) into tier 1 — Simplified Chinese, Japanese, Korean. Budget $100-300 per language.
- Set localized app names for those three languages so local search matches what players type.
- Watch regional traffic in Steamworks for 4-8 weeks. You are looking for a mix shift: Chinese and Korean impressions, visits, and wishlist adds growing while English regions stay flat.
- Whichever languages move, those are the ones that earn full in-game localization at $0.09-0.14 a word.
This is the cheapest market research on Steam. Three page translations cost less than most capsule art revisions and produce region-by-region demand data specific to your game, not to the platform average.
Localization and regional pricing are one decision
The languages in tiers 1-2 overlap heavily with Steam’s discounted pricing regions, and the two levers only work together. A Chinese player who can finally read your page still bounces if you lazily charged the full US price in CNY; a perfectly priced RUB or TRY entry does nothing if the page is in English.
Political Animals is the clean illustration: units up 3.2x, revenue up meaningfully less, because Valve’s recommended China price at the time was around half the US price. That trade is usually worth taking — those are sales that did not exist before, plus review volume that feeds visibility everywhere — but you should price it deliberately. Our cheapest Steam regions breakdown shows how deep the discounts run by country, and the regional pricing calculator converts your USD price into Valve-recommended regional prices so you can model revenue per localized market before you commission a single word.
Budget both lines at once: if you localize into Simplified Chinese, Russian, Turkish, and Brazilian Portuguese, you are also implicitly deciding your price in four of Steam’s most price-sensitive regions.
How to pick your set: a decision rule
Work through this in order:
- Count your words. Export every string. Under 10,000 words, plan for 8-10 languages; over 50,000, plan for 3-5 and choose hard.
- Buy tier 1 store pages now. Simplified Chinese, Japanese, Korean page translations, ~$300-900 total, regardless of what you decide about the full game.
- Read your own regional data. Steamworks traffic by country plus wishlist regions. A strategy game with German traction or a horror game with a Japanese streamer bump should reorder the default tiers — your genre has its own geography.
- Fund full localization from the test results. Languages whose store page moved traffic get in-game text at market rates. Languages that stayed flat get skipped without guilt, whatever any generic tier list (including this one) says.
- Set the language checkboxes the day each translation ships. Supported-language flags gate your visibility in each region’s browsing surfaces; an unticked box hides you from that cohort no matter how good the translation is.
- Retranslate the page once before launch. You will rewrite your English copy; budget one final pass per language after it locks, or your Japanese page will describe a game that no longer exists.
The one-sentence version: Simplified Chinese, Japanese, and Korean pages this month for a few hundred dollars, full localization only where your own regional numbers say the demand is real.