For the past few years, there’s been this viral idea that translation is essentially a technical problem. Text goes in, then text comes out. The rest is just computing power.

Yet a recent article from the FAZ entitled Wie Übersetzer es mit Künstlicher Intelligenz aufnehmen (How Translators Are Responding to Artificial Intelligence) reads like a necessary correction to that way of thinking. Not because AI is failing, but because it is often misunderstood.

In practice, AI and machine translation tools have long been part of everyday professional workflows. They speed things up, improve consistency, and help with routine tasks. Anyone working seriously with language who ignores these tools is working inefficiently. It’s that simple.

Yet Speed Doesn’t Replace Responsibility

It is precisely in specialist, strategic, or legally sensitive texts that the blind spots of the automation narrative become visible. According to Ekaterina Lapshinova-Koltunski of the University of Hildesheim, AI-generated translations are still very literal compared to human translations… very close to the source text. With generative AI, there is an additional risk of hallucinations, meaning content that is simply made up.

This is Not a Theoretical Problem

A text might be linguistically correct, but can still be factually imprecise, the tone can be off, or the context can be risky. And it’s exactly these kinds of deviations that become expensive for companies later on.

The decisive question, then, is not whether AI will replace translators.
The question is: who is evaluating, steering and assuming responsibility for the output?

Professional translation work today is often a matter of contextualizing AI output: checking terminology, recognizing implicit meaning, controlling tone and effect. In short, knowing when a text works, and when it doesn’t.

For companies that already use AI, or are considering it, there’s no contradiction here.
Efficiency and quality are not mutually exclusive. But trust doesn’t emerge automatically. It requires expertise.

Language is not a neutral delivery mechanism. So it’s shouldn’t be automated without oversight.

To learn more about how I combine AI with oversight, check out my AI & Automation services here.

🇩🇪 Lesen Sie hier den Beitrag auf Deutsch / 🇫🇷 Lisez l’article ici en français