Definition
Software that automatically converts text or speech from one language into another, with no human translating by hand.
At a glance
- Modern systems use neural machine translation: they learn from millions of human-translated examples and aim at meaning, not word-for-word swaps[1].
- Fast and cheap, so it is practical for bulk content like product listings, support tickets, and emails.
- Strong on common languages and everyday text; weak on idioms, tone, numbers, and fields like legal, medical, or financial.
- Match the workflow to the stakes: machine-only for low-risk volume, human review for anything affecting trust or compliance.
How it works
Tools like Google Translate and DeepL feed whole sentences through large neural networks trained on translated text, then produce natural-sounding output[1]. The same engines plug into your website, help desk, or apps.
Where it stumbles
It garbles idioms, brand voice, and exact details like numbers or dates[3]. In regulated areas, a confident but wrong translation can create real legal liability[4], and the systems rarely flag their own mistakes.
Why it matters
The market was near 1.1 billion dollars in 2025 and is growing at double-digit rates[2]. A practical setup is tiered: machine alone for bulk low-risk material, light human post-editing for important content, full human translation for high-stakes documents[3].
Bottom line
Treat machine translation as a powerful first draft: let it carry the volume, and keep a human on the few items where being wrong is costly.