A taco shop in Houston where half the regulars order in Spanish. A Montreal bakery that lives in French and English. A Berlin SaaS selling to Spain and Mexico. Sooner or later every one of them asks the same question: which language should our social media be in? And the usual answer is “all of them.” After that, posts come out in one language or the other, the Spanish version shows up two days after the English one, and half the caption has clearly been through a translation app.
Having several languages isn’t the problem. The problem is that the second language gets whatever time is left over: write the post, then rush a translation. You end up doing the work twice, and a native speaker still spots the translation in the first line.
Below: how to pick a setup, what to translate versus rewrite, a one-page glossary, and a workflow where the second language costs 10–15 minutes per post instead of another workday.
Three setups and how to choose
| Setup | What it looks like | Works when | Downsides |
|---|---|---|---|
| Two languages in one post | English caption, Spanish below it | Short posts: announcements, hours, promos | Long captions; the second language hides under “more” |
| Alternating in one account | Monday in English, Tuesday in Spanish | Audiences overlap; the second language is under a third | Every follower skips half the feed |
| One account or channel per language | @brand and @brand_es | Different audiences, capacity for two plans | Two follower bases, double support load |
How to decide in practice:
- Look at numbers, not gut feeling. Instagram insights break your audience down by city and country, and comments and DMs tell you the language. If 85% of people message you in English, a separate Spanish account will sit empty.
- Two languages in one post hit the limits. An Instagram caption can run 2,200 characters, but only about 125 show before “more.” So the first line is always in one language, and the other reader sees text they skip. On Telegram, a photo caption is capped at 1,024 characters — two full paragraphs in two languages won’t fit.
- Telegram almost always means separate channels. Subscribers get a notification for every post, and posts in a language they don’t read are the fastest route to unsubscribing. Channel names and descriptions show up in global search, so “Custom furniture | Houston” and “Muebles a medida | Houston” catch different queries.
- YouTube has this built in. In Studio you can add translated titles and descriptions: viewers see them in their interface language while the video stays one video.
Auto-translate doesn’t replace a real version. Instagram and Threads show “See translation,” and between English and Spanish it’s decent — but having to tap it tells the reader the post wasn’t written for them.
What to translate, what to adapt, what to rewrite
Not all posts are equal. Sort your content into three buckets and the second language takes half the time.
- Translate: facts. Hours, addresses, prices, delivery terms, promo announcements. Accuracy matters more than style here, and a machine draft plus a human check does the job.
- Adapt: useful content. How-tos, breakdowns, checklists. Same idea, but the examples, units and references change. A “what to check before signing a lease” checklist reads differently for a long-time resident and for someone who just moved from Monterrey — they have different questions.
- Rewrite or don’t carry over: voice. Humor, opinions, reactive posts, wordplay. A “case of the Mondays” joke doesn’t survive in Spanish. Either someone who thinks in that language writes these, or they stay in the main version only.
An adaptation, not a translation. Core message: “from October 1 we open at 8 am.”
- en: “New hours from October 1: doors open at 8 am. Grab a cold brew before I-10 turns into a parking lot.”
- es: “Desde el 1 de octubre abrimos a las 8 a. m. Pasa por tu cafecito antes de entrar a la chamba.”
- en, for visitors: “Open 8 am daily from Oct 1. Card and Apple Pay accepted, Wi-Fi password at the counter.”
One fact, three versions, each speaking to what its reader cares about. A visitor cares more about payment than traffic.
Glossary and voice: one page per language
The main reason second versions read badly: the same term gets translated differently every time. The fix is a one-page glossary for each language:
- What we never translate. Brand, product and plan names, platform terms: Reels, Stories, engagement rate.
- How we address the reader. Spanish makes you choose “tú” or “usted” (and “vos” for Argentina), French “tu” or “vous,” German “du” or “Sie.” Decide once and stop debating it in every post.
- Script and variant. Chinese has Simplified (mainland, Singapore) and Traditional (Taiwan, Hong Kong); Serbian runs in both Latin and Cyrillic; Spanish differs between Spain and Latin America. Pick one per account and stick to it.
- Accents and punctuation in hashtags. #café and #cafe are different hashtags, and an apostrophe ends one: #mom’skitchen becomes #mom. Decide on one spelling and use it everywhere.
- A 10/10 word list. Ten “we say” and ten “we don’t say” words in each language.
Brand voice doesn’t translate word for word either: tone scales and reference posts are set up separately for each language. We covered how to build that document in the article on tone of voice. Keep it in the same place as your product facts — a brand knowledge base — so people and AI pull terms from one source.
Workflow: the second language in 15 minutes
A routine that holds up at five posts a week in two or three languages:
- Core (5 minutes). Point, facts, action — three lines. Same logic as cross-posting: you adapt the message, not the text.
- Main version (15–20 minutes). In the language most of your customers use. Written at full strength.
- Draft of the second version (2 minutes). Give the model the core, the glossary and 2–3 reference posts in the target language — not the finished English text. Prompt: “Write a post in Mexican Spanish from this core. Use ‘tú’, take terms from the glossary. Don’t translate the English version; write for a reader in Houston.”
- Native review (5–10 minutes). Not optional. You’re looking for three things: calques from English, a wrong term, an unnatural form of address. No native speaker on the team? Agree on a review package with a freelancer — it’s cheaper than unfollows.
- Publishing. If both audiences share a time zone, space the versions 1–2 hours apart so people who follow both accounts don’t see the same post twice in a row. If the languages are different countries, schedule for each local best time: New York and Los Angeles are three hours apart, Madrid and Mexico City seven or eight.
How to measure results by language
Don’t roll everything into one number. Track reach, saves, clicks and inquiries for each language separately and compare them to the time spent. The typical picture after three months: one language brings most of the leads, a second builds awareness, a third does almost nothing. That’s not failure, it’s information — cut the third down to translated facts.
The biggest mistake is judging after two weeks: a new account in a second language needs 6–8 weeks of consistent posting.
When you don’t need multiple languages
- 90% of your audience reads one language. A second one “to look international” eats hours and brings no customers.
- Nobody can review. Machine translation without a native check is worse than no version at all: it signals you don’t care about that language.
- Support only speaks one language. If the post is in Spanish but DMs are answered only in English, you’re promising something you don’t deliver.
- Regulated industries. Healthcare, finance, legal: a mistranslated condition is a legal risk. A professional translates those, not a model.
With two or three languages, most of the effort goes not into writing but into keeping track of which version goes where and when. In Repost21 your Spanish Instagram and your English Facebook Page live in one content calendar, the glossary and voice connect live from Google Docs or Notion as a brand knowledge base, and Luna drafts posts based on them. Try it free.