Somebody watched your service, understood some of it, and sent a message in a language nobody on your staff reads.
For most of church history the answer to that was entirely a matter of luck: you could serve a language if a bilingual person happened to be in your congregation, and you could not if one did not. No amount of conviction changed it. That is a hard ceiling, and it is the one machine translation has genuinely removed.
It is the clearest multiplication case in this series. It is also the one where the limits matter most, so both halves are worth saying carefully.
What actually changed
Machine translation got good in a way that is easy to underestimate if you last tried it in 2015. For ordinary prose between major languages it is now roughly as good as a competent non-specialist human, instantly, at no meaningful cost.
That makes several things possible for a church that were not possible three years ago:
- Answering a message in the language it arrived in, the same day, without a volunteer who happens to be available.
- Practical information — service times, what happens when you arrive, how to ask for prayer — available in whatever languages are actually showing up.
- Getting someone to the right resource, in a language they read, rather than apologising for not having one.
And it changes the question a small church asks. It is no longer “do we have someone who speaks Portuguese?” It is “who is already here that we have been failing to answer?” — which is answerable from your own messages and analytics, and is usually surprising.
Heart language is the reason this is not a courtesy
There is a distinction worth holding onto, because it explains why this is ministry rather than customer service.
The language someone can do business in is often not the language they pray in. People follow an argument perfectly well in a second language and still find that it stays at arm’s length — that it is information they understood rather than something that reached them. Heart language is the one you think in, grieve in, and were loved in as a child. TCD has a whole guide on evangelism in heart language and the short version is that the gap between comprehension and reception is exactly where this matters.
Which means translating your ministry is not about being accommodating. For a lot of people it is the difference between having heard something and having received it.
Where machine translation should not be trusted
Now the other half, and it is not a small caveat.
Machine translation is least reliable precisely where churches most want to use it: Scripture, doctrine, and the vocabulary of faith. The reason is structural rather than a matter of quality improving over time. Words like grace, repentance, covenant, spirit and even church have technical meanings that were settled by centuries of theological work in each receiving language, and translators of Scripture made deliberate, contested choices about them. A general-purpose model does not know your tradition made those choices, and will confidently produce a word that means something adjacent and wrong.
The dangerous part is that the output is fluent. A clumsy mistranslation announces itself. A smooth one gets read and believed, and nobody in the room can catch it, because if anyone could read that language you would not have needed the machine.
So the practical line I would draw:
- Use it freely for practical, pastoral and conversational text — information, first replies, getting someone to the right place.
- Do not generate Scripture with it. Link an existing published translation in that language. Bible translation is somebody’s life’s work and it already exists; this is not the place to improvise.
- Get a human speaker before anything doctrinal or personal, and say plainly that translation is being used. Most people are entirely fine with that, and hiding it is what turns a reasonable tool into a breach of trust when it surfaces.
Where this ends up
The goal is not a translated website. It is a person at the other end who speaks that language.
Machine translation is what makes the introduction possible — it gets you from “we cannot answer this at all” to “we can answer this today and find someone properly.” That is an enormous change for a small church, and it is still an introduction rather than a relationship, which is the same line this series keeps arriving at from different directions.
Start by finding out which languages are already in your messages. Not which ones you imagine you should serve — which ones are already showing up and going unanswered. For most churches that list is short, specific, and has been sitting in the inbox for a while, and the person you hand it to can now actually do something about it.
Part of the weekly AI series. The full guide is AI for churches.
