In this article
A channel that only publishes in one language is read by people who read that language, which sounds obvious and is usually discovered late — typically when the statistics show a large audience in places whose language you do not speak.
There are four ways to answer that, and they differ enormously in what they cost you. The deciding question is not which languages, it is whether anybody on your side can read the replies, because publishing in a language is easy and being answerable in one is not.
What you will be able to do after this step
- check_circle Choose between four arrangements on their real costs.
- check_circle Say why the second language in a bilingual post is always read second.
- check_circle Avoid the dead regional channel that is worse than none.
- check_circle Handle right-to-left text and time zones without guessing.
Four arrangements, four different bills
Each of these works for somebody. None of them works for everybody, and picking by ambition rather than by capacity is the usual mistake.
Publish in one language and leave it to the reader
Telegram offers readers a translation control of its own, covered in Step 053. It costs you nothing, it is already there, and for a large part of your international audience it is what they are quietly doing already.
It works best where the content is factual and the sentences are plain: announcements, instructions, numbers with context around them. It fails where meaning lives in tone — a joke, a careful qualification, an apology — because register is the first thing machine translation loses and the reader has no way to know it happened.
What you can do to help, without translating anything, is write for it: short sentences, few idioms, names and numbers stated plainly, and important qualifications inside the same sentence rather than left to a following one. That is better writing for everybody, and it survives translation far better than the version with the clever phrasing.
The correct default for almost every channel under a few thousand readers. Do nothing, write more plainly, and revisit when the statistics say otherwise.
One post carrying two languages
English at the top, a divider, the second language beneath. It keeps one channel, one audience and one schedule, which is a real advantage, and it has one structural flaw that no formatting fixes.
The second language is below the fold, every time. On a phone the post opens at the top, so one of your two audiences reads first and the other scrolls past a wall of text that is not for them to reach theirs. That asymmetry is small per post and cumulative over months: the group placed second learns that the channel is not really theirs.
It also doubles the height of everything, which changes how the channel feels to scroll, and it stops working entirely at three languages — nobody scrolls past two blocks to reach a third.
If you do this, alternate which language goes first, or put the shorter one on top. And keep the posts short, because doubled length is the cost you are paying for simplicity.
Fine for exactly two languages and a low posting rate. Collapses at three, and one audience is always reading second.
A channel for each language
This is the arrangement that actually works beyond two languages, and the cost is not translation. It is that you now run several channels, each with its own posting schedule, its own comment room (Step 075), its own moderation and its own silence when nobody is looking after it.
The failure mode is a dead channel, and a dead channel is worse than none at all. A reader who finds a Spanish channel with nothing since the spring learns that you started something for them and stopped. You would have been better never opening it, and the decision to open it was made at a moment of enthusiasm by somebody who was not going to be the one posting in October.
So the test before creating one is a person, not a market: who writes this every week, who reads the comments, and what happens when they go on holiday. If the honest answer is "a tool will", you are on the fourth tab and should read it carefully.
Where it goes well it goes very well — independent schedules mean independent time zones, native discussion rooms, and partners in that region who can be approached properly. Pin a small directory in every channel linking the others, so that a reader who lands in the wrong one can move.
The only arrangement that scales past two, and the only one with a staffing requirement. Open one per person you actually have.
Automatic translation into regional channels
A bot watches the main channel and republishes everything, translated, into the others. It is cheap, it is fast, and it is visibly what it is.
The problem is not the quality of the translation, which is often decent. It is that nobody is reading the output before it is published. Machine translation fails on exactly the things a channel gets judged on — names, numbers, negations, and anything where politeness is encoded in the grammar — and it fails fluently, producing a confident sentence that says the opposite of what you wrote. In a language you cannot read, you will not find out until somebody tells you, and most people will not tell you. They will just stop reading.
The version that is defensible is machine translation plus a person who reads it before it goes out. That is a good use of a tool: it removes the typing and keeps the judgement. It is also no longer automation, and it should not be sold to you as such.
One more thing to decide deliberately: an automated relay that reposts instantly wakes the other regions at your working hours. Send the relayed posts silently, or schedule them.
Unattended machine translation publishes mistakes under your name in a language you cannot check. Use the tool, keep the reader.
forum The decision is about answering, not publishing
Publishing in a language is a translation problem and it is solvable by the hour. Being present in a language is a different commitment, and it is the one that determines whether a regional channel is an asset or an embarrassment.
Open a channel and you open a comment room with it. Somebody has to read that room daily in that language: to answer questions, to notice when a new member is being approached by a scammer in a language you do not speak, and to see the tone change before it becomes a problem. Step 061's automatic filtering does not understand your subject in any language, and moderation you cannot read is moderation you are not doing.
There is a middle position worth considering more often than it is: publish in one language, and appoint one person per region who is willing to answer. No second channel, no relay, just a named human in the comments who speaks it. It handles most of the real problem at a fraction of the cost, and it is reversible.
schedule Time zones, and posting into somebody's night
A single global channel has no correct posting time, so it will inevitably ring at three in the morning for a meaningful slice of the audience. The three tools from earlier steps do the work here.
Schedule rather than publish (Step 073), so a post written at your convenience arrives at theirs. Send Without Sound (Step 074) for anything that does not need to wake anybody, which is most of it. And if you do run regional channels, the whole point is that each one has its own clock — a relay that fires everything simultaneously throws that advantage away on the first day.
The thing to measure is not reach but muting. Waking people at night is the fastest way to be silenced permanently, and a channel that has been muted looks exactly like a channel that is being read.
format_align_center Right-to-left text, and why it looks scrambled
Arabic, Hebrew, Persian and Urdu run right to left, while Latin words, digits and most punctuation run left to right. When both appear in one sentence the software has to decide where each fragment sits, and around brackets, hyphens and quotation marks it frequently decides differently from what you intended. The symptom is a line where a price, a name or a URL has apparently jumped to the wrong end.
Three practical remedies, none of which requires you to understand the rules. Keep Latin fragments on their own line rather than embedded mid-sentence. Avoid punctuation at the boundary between the two scripts, which is where nearly all of it goes wrong. And have somebody who reads the language look at the rendered post, on a phone, before it is published — not at the text you typed, which will look correct to you in either case.
The same instinct applies to layout generally. Interfaces mirror in right-to-left languages, so anything you describe by position — "the button on the left" — is wrong for those readers. Describe controls by their name or icon instead, which is better writing anyway.
touch_app Try it now — five minutes
Four steps, and the first one tells you whether the rest of this page is your problem yet.
- Open your channel's statistics and look at languages and regions If there is no meaningful audience outside your own language, nothing here is a decision you need to make today.
- Translate one of your own posts and read it back Use the reader's translation control on your own writing. What survives and what does not is the clearest guide to how to write for a translated audience.
- Before opening any regional channel, name the person Not the tool. Who posts and who reads the comments, and what happens in August.
- Schedule your next post for the other side of the world's morning One post, silently. It is the cheapest test of whether the audience there is real.
help Common questions at this stage
Should the handle say which language it is?
Yes, and put it at the end rather than the beginning so the brand reads first. A reader arriving from a search or a forwarded post should be able to tell in a glance whether this is their channel, and a handle that requires opening the channel to find out costs you exactly the people you made it for. Keep the naming pattern identical across all of them, and remember that a public handle is not easily changed once printed anywhere.
Can I translate a post after it has already gone out?
You can edit a channel post indefinitely, so adding a second language later is possible — but editing is silent, as Step 081 notes, so nobody who already read it will be told. For a bilingual channel the practical answer is to publish complete or to publish twice. Editing a post to add a language an hour later reaches only the people who had not read it yet, which is the audience you were least worried about.
Does a separate channel per language split my subscriber count?
It splits the number and usually grows the total, because a channel in somebody's own language is one they will actually read and recommend. The number to stop watching is the headline count on any single channel; the number that matters is whether each channel is read. A large multilingual channel where most subscribers have muted it is worth less than three smaller ones people open.
What about the discussion group — one shared or one each?
One each, if you have regional channels at all. A shared room where four languages are being spoken at once is unreadable for everyone in it, and it is the single most common reason a multilingual community feels chaotic. If you cannot staff separate rooms, it is a strong argument for not splitting the channels either — the room is where the cost of internationalisation actually lands.
How do I keep product names from being translated?
You cannot control what a reader's own translation does, so the defence is in how you write. Names that look like ordinary words get translated; names that are distinctive do not, and neither do names written in a consistent form the reader has seen before. Where a term genuinely must not change, the reliable trick is to include the original in brackets the first time it appears, which survives translation because the bracketed fragment tends to be carried across intact.
Is it worth translating the channel description and pinned post first?
Yes, and if you only do one thing, do this. The description and the pinned message are what a stranger reads before deciding whether to stay, and they are two short pieces of text that do not change often — which makes them the highest return on a human translation you will ever get. A channel whose posts are machine-translated but whose introduction is written properly reads as a channel that made an effort. The reverse reads as neglect.
At a glance
- 1Four set-ups
Readers translate, one post in two languages, separate channels, or a bot.
- 2Second language last
In a bilingual post one language is always read second.
- 3Dead channels hurt
Open a language channel only if someone will answer in it.
Updated
