Why translation isn’t enough. And why localisation almost always is.

Most international learning projects begin with a budget assumption that quietly causes problems later.

The assumption is that translation is the main cost. Move the content from one language to another, get it signed off, repeat for each target country. Job done.

Translation is part of the cost. It’s rarely the most important part. And in my experience of delivering learning in 38 languages over nearly four decades, translation alone is almost never enough to make learning land properly with the people who need it.

The work that makes the real difference is localisation. And it’s the bit that gets underestimated, undervalued, and underpaid more often than any other element of international learning design.

This is what I want to talk about.

Translation moves words. Localisation moves meaning.

Translation does what it says. It takes content in one language and converts it into another. Done well, it’s accurate, grammatical, and faithful to the source.

Localisation goes further. It considers everything that surrounds the words. The examples. The names of the characters in scenarios. The imagery and stock photos. The hierarchy implied by how the reader is addressed. The level of formality. The use of humour. The cultural assumptions baked into the original content that don’t travel.

A piece of learning that has been translated will be understood. A piece of learning that has been properly localised will land.

Those are two very different outcomes, and the difference shows up in completion rates, application rates, and ultimately in whether the learning achieves what the business wanted from it.

Why AI translation isn’t the threat it’s being sold as

AI translation has improved dramatically in the last few years. Modern tools produce results in seconds that are grammatically sound, contextually reasonable, and frequently usable as a starting point.

I use AI translation tools myself. They’re genuinely useful, particularly in the first pass of a long document, in tight budgets, or in conversational situations.

But “starting point” is the right phrase. Because what AI produces is translation, not localisation. And in learning content specifically, the gap between those two is where the value lives.

AI can’t read the room. It doesn’t know which examples will resonate in São Paulo versus Lisbon, even though they share a language. It doesn’t know that a joke that works in Madrid might fall flat in Bogotá. It doesn’t know which words have drifted in meaning over the last few years. It can’t spot when a brand name in one market is unintentionally rude in another.

A practical approach I’ve seen work well is to use AI for the first pass, get the bulk of the translation done quickly and cheaply, and then bring in a real local person to localise that draft. Their job at that point is not to translate again. It’s to identify what AI got technically right but culturally wrong, and to adjust accordingly.

That hybrid approach is dramatically better than AI alone, and significantly cheaper than full traditional localisation done from scratch.

Real local people in real local territories

When I work with localisation companies, I look for one specific thing.

Real local people, living in the local territories.

Not someone who studied the language at university. Not an expat who’s been away for ten years. People who are currently embedded in the culture, who hear the language of the streets, who know what’s trending in conversation, who can spot when a word has shifted meaning or when an idiom has fallen out of fashion.

That’s the only kind of localisation worth paying for. Anything less is just translation with extra steps.

An example of why this matters. A French Canadian who has lived in France for ten years is sometimes asked to localise content into Canadian French. Their Canadian French, however, is now ten years out of date. It has been mixed with European expressions they have absorbed without noticing, and influenced by cultural references that have shifted while they have been away.

The result is technically Canadian French. It just doesn’t sound quite right to actual French Canadians.

Distance from a culture costs you accuracy, even when you speak the language fluently. This is true across every language pair you can think of.

The hidden value of doing learning twice

A story from a project I worked on, delivering learning in 11 languages for a single client.

The client had a points-based incentive system. Complete the learning, earn points, spend the points on merchandise and product discounts. A standard approach to encouraging engagement.

After a few months, the client noticed something they didn’t like. Some learners were completing the learning in more than one language, usually English plus their native language. The client flagged it as exploitation of the points system and asked whether they should restrict it.

I gently disagreed.

Yes, those learners were earning more points than the system had intended. But they were also absorbing the content twice. Picking up perhaps 80% of it in English first, then reinforcing 100% of it in their native language. They were not gaming the system. They were learning more effectively than anyone else in the cohort.

The client took the point. The behaviour wasn’t a problem. It was a feature.

That story illustrates something important about international learning. When you’ve genuinely localised the content and given learners a choice, they will often choose paths that improve their own learning in ways nobody planned. Treating that as a problem misses the whole point.

The plain truth is that anyone doing learning more than once is doing better than anyone not doing it at all.

Plain speaking matters even more internationally

When you’re writing for an international audience, plain speaking matters more, not less.

If your audience is reading or hearing your content in their second or third language, every unnecessary complication in the source material gets multiplied. Idioms become confusing. Jargon becomes impenetrable. Clever wordplay becomes irritating. The structure you thought was sophisticated becomes a barrier to understanding.

The clearer the original, the better the localisation. The simpler the structure, the more chance the content has of landing well across cultures.

International work has, slightly to my surprise, made me a better writer of English. It forces you to strip away everything you don’t actually need.

A budget reality worth knowing

Most international learning projects underestimate localisation costs.

Translation tends to be linear. So many words, so much money, done. Localisation is iterative. It involves cultural review, expert feedback, adjustments to imagery, examples that need replacing, and tone that needs calibrating for each audience.

Done properly, localisation can double the project budget compared to translation alone. It also doubles or triples the chance of the learning actually working in the target markets.

If you’re scoping an international learning project, plan for that. It’s far cheaper to get the budget right at the start than to discover six months later that the content hasn’t landed and needs to be redone.

A summary, plainly

Translation is the technical bit. Localisation is what makes learning actually land.

Real local people in real local territories will always beat clever technology working from a distance.

AI is a useful first pass. It is not a substitute for human cultural judgement.

Plain speaking matters more, not less, when your audience is reading in their second language.

And anyone doing the learning more than once is doing better than anyone not doing it at all.