Google Translate is fluent now. It still isn't native.

You are at a tiny counter three streets from anywhere a tour bus goes. The owner is asking you something, smiling, waiting. A year ago you would have typed a reply and turned your phone around. Today you might just hold up Google Translate and let it answer for you, out loud, in Japanese, a few seconds behind. That is a genuine upgrade, and it deserves to be said plainly: the words will be correct, and they will be spoken in a natural voice.

So if the words are right, what is left to get wrong?

Fluent is not the same as native

Quite a lot, as anyone who has lived in a second language will tell you. The best machine translation today is astonishingly fluent. It is also, still, not native. It speaks like a brilliant foreigner who studied for ten years: grammar flawless, vocabulary enormous, and every so often a phrasing a person who grew up there would simply never choose. Correct, and a little bit off. You can feel the English underneath it.

Linguists even have a name for it: translationese. The fake fluency of a sentence that is grammatically perfect and still reads like a translation rather than something a person would actually say. It is the exact thing serious organizations work to scrub out, the OECD warns against a “translationese flavour” in its international surveys precisely because a technically correct translation can quietly distort the meaning. It is subtle, and it is everywhere machine translation is.

The reason is that a general translator is solving one problem, turn these words into accurate words in another language, and it solves it at a flat, default politeness. But “say this in Japanese” has no single answer. It has a dozen correct answers, and a local picks between them without thinking, based on who is in front of them.

Where the gap shows

It shows exactly where it matters most: the people, not the signs.

• In Japan, keigo runs from casual to honorific. A perfectly translated sentence in plain form, said to someone you have just met, lands as blunt even though every word is right.
• In Thailand, whether you end a sentence with khrap or kha, and the deference an elder is owed, is not decoration. It is the message.
• In Spanish, tú or usted is the difference between warm and presumptuous, and a default setting will quietly pick wrong.

None of these are vocabulary problems. They are register, idiom, and the hundred small choices a native makes that a fluent foreigner, human or machine, tends to miss.

It loses the thread

There is a second gap, and it has the same root: a fast, fluent translator works one sentence at a time. Real conversation does not. We lean on what was just said. “Can I bring my wife?” and a beat later, “She’s vegetarian, by the way.” A person hears “she” and knows. A tool that translates each sentence on its own has already let go of the wife, so “she” arrives as a guess.

Natives talk in shorthand because they trust you to hold the thread. “The same again.” “Him too.” “Like last time.” Take the memory away and you are forced to over-explain, restating context a local would never repeat, which is its own quiet way of sounding foreign. Something that follows the whole conversation, not just the line in front of it, keeps the thread the way the person across from you does, so “she” stays your wife and “too” still means coming to dinner.

Why a focused tool can close it

This is the case for doing a few languages properly instead of a hundred passably. A translator that supports seventy languages is, by necessity, fluent in all of them and native in none. A tool built for a short list can do the slower work: choosing the register for the relationship, using the phrase a local would actually use, and getting the dialect right, Mexico or Madrid, Northern or Southern, casual or formal.

That is what it means to interpret rather than translate. Not just convert your sentence, but decide how to say it, for the person in front of you, the way they would say it themselves.

Which one do you reach for?

Reach for Google Translate for what it is genuinely great at, and now a little more: reading the world, a quick solo exchange, a fast spoken back and forth where nobody’s feelings are on the line. It is free and it is excellent, and you should keep it.

When there is a person you actually want to be understood by, your partner’s family, an elder, a host who is being kind, the bar moves. You no longer want the merely correct sentence. You want the one a local would have chosen, with the respect the moment asks for. Fluent gets you understood. Native gets you in.

The next time it is a person who matters, meet the interpreter.