Neural machine translation has improved dramatically. For well-resourced language pairs — English to Spanish, French, or German — the output reads naturally enough that you might be tempted to ship it without a second look. And for internal, low-stakes communication, that's often fine.
But the moment translated text faces your customers, your regulators, or your brand, "good enough" starts showing its edges. Here's where MT quietly breaks — and why a native linguist still has to catch it.
Where MT fails silently
The dangerous thing about modern MT isn't that it produces gibberish. It doesn't. The dangerous thing is that it produces text that looks correct but carries the wrong meaning in ways that only a native speaker would notice.
1. Idiom and cultural register
MT translates the words, not the intent. A marketing tagline like "We've got your back" might render literally in German or Japanese, losing the reassuring connotation entirely. A native linguist recognizes the idiom and finds the cultural equivalent — not a word-for-word mirror.
2. Legal and regulatory weight
In contracts, terms of service, and compliance documents, word choice isn't creative — it's legally binding. MT doesn't know that "shall" and "will" carry different obligations in English legal prose, or that the equivalent distinction in French contract law uses different formulations entirely.
MT output is never sent as final copy. It's a first draft — and first drafts exist to be edited.
3. Tone-of-voice consistency
Your brand voice is specific. MT doesn't read your style guide. It won't know that your product uses informal "you" in German (du vs. Sie), or that your help center avoids passive voice. A post-editor catches these and aligns every string to your existing tone.
The economics of MTPE
Machine Translation Post-Editing (MTPE) isn't a compromise — it's a strategy. Here's why it makes sense:
- Speed: MT produces a first draft in seconds. The linguist spends time refining, not starting from scratch.
- Cost: MTPE typically runs 30–40% less than full human translation for suitable content types.
- Consistency: When combined with a translation memory and glossary, MTPE can actually improve consistency across large volumes.
The key is knowing which content types are suitable for MTPE (help articles, support tickets, internal documentation) and which need full human translation from the start (marketing, legal, medical).
What a good post-editing pass actually checks
A trained post-editor does more than proofread. They:
- Verify meaning accuracy against the source text
- Correct terminology against the project glossary
- Fix register and tone to match the style guide
- Check formatting: placeholders, tags, length constraints
- Flag segments where MT output should be discarded and rewritten
This is why we never ship MT output as final copy. Every string that leaves Mavy EduSolutions has been reviewed by a native linguist — whether the first draft came from a human or a machine.
Need a quote for MTPE or full human translation? Start a project — tell us your content type, word count, and target languages, and we'll recommend the right tier.