Skip to content
Faisal Mahmud TusherDigital marketing, SEO & analytics
with Faisal Tusher

Glossary term

Machine Translation

Translation produced by software or a language model, now fluent enough for first drafts but used with human post-editing applied by content type.
StrategytechnologyUpdated

Definition

Machine translation, often shortened to MT, is translation generated automatically, whether by a dedicated translation engine or by a general language model asked to translate. The output is a draft. When a qualified linguist then edits that draft against the source, the termbase and the style guide, the work is called post-editing, and the combination of the two is the model the localisation guides describe.

It differs from the other ways copy reaches a new language by where the effort goes. Human translation puts the effort into converting the words faithfully; transcreation puts it into rewriting the idea from the brief; machine translation removes most of the effort from the first draft and moves it to review.

Where it helps and where it does not

The guide to translation and transcreation sorts copy by type. Product data, help articles and long informational content are where a machine draft, constrained by the termbase and fully post-edited, saves the most. Web pages and emails can use it as a first draft or a reference, edited to the style guide with every claim and term checked. Headlines, ads, taglines and campaign concepts are not a good starting point, because the work there is rethinking the idea, and legal, regulatory and claims text goes to professional translation with local compliance review.

The same guide names where errors cluster: brand and product terms, claims, tone, idiom and anything that depends on context the model was not given. Those are also the places marketing is most exposed. The international SEO hub adds a commercial test: a market that would receive a machine translation and nothing else is usually better served by the primary edition with clear shipping information, and automatic translation published for search without review can be treated as auto-generated content.

In practice: post-editing by content type and in-country review

Three habits keep the method honest. The localisation brief template asks the brief to say whether machine translation is used and for which assets, so nobody discovers it at review. The guide to choosing and managing translation suppliers asks each supplier to state in writing where it is used in their process and who post-edits it, so the work delivered matches the label agreed. And the editorial controls that apply to any AI-assisted copy apply here, with one addition: the reviewer must be fluent in the target language, because a reviewer who cannot read the output is checking layout, not meaning. The localisation hub places the method inside the wider workflow of briefs, review and sign-off.

Illustrative example — not a real result

A board game publisher splits rulebooks from campaign copy

Steangulm Board Games prepared French and German editions of 3 games: 11 rulebooks totalling about 38,000 words, and a launch campaign of 26 ad headlines, 6 emails and a landing page. It ran everything through machine translation. Post-editors working from a 310-term glossary of card and component names corrected the rulebooks at roughly half the cost quoted for translation from scratch, and an in-country review found 2 rules that needed rewording. The campaign drafts were accurate but flat: a pun in the main headline vanished in both languages. The publisher sent the headlines and emails to a native copywriter in each market with the original brief, and kept the machine drafts only as a reference.

Transcreation is the method for the copy machine translation handles worst, rewriting the idea from the brief. In-Country Review is the check by a fluent reviewer in the market that every machine draft still needs. A Translation Memory keeps the approved translation of each segment so repeated wording comes back the same way; it reuses text a reviewer approved rather than generating new text. Back-Translation translates the target text back into the source language to check what risky lines say; it checks meaning, and a fluent back-translation can hide problems as easily as a fluent machine draft.

Common confusions

  • Fluent output is accurate output. Fluency hides errors; they are found by customers if no qualified reviewer looks first.
  • Machine translation and transcreation are alternatives for the same copy. They suit different content types; persuasive lines go to transcreation.
  • Post-editing is proofreading. A post-edit checks the draft against the source, the termbase and the style guide, not only for typos.
  • A supplier's price tells you whether MT was used. Ask in writing where it is used and who edits it.
  • Any version is better than none. For a market that would get machine output alone, the primary edition is often the better answer.

More terms to look up

Where this term is used

Related