DeepL or Google Translate: Which one mangles your contracts, PDFs and emails less?

You have a ten-page lease in English and a landlord waiting for your signature. You paste it into a free translator, and the clauses come back readable — until you reach the table of payment dates, which has collapsed into a wall of numbers. Nothing is technically wrong with the words. The shape of the document is gone.

That gap between readable words and a usable document is where the two biggest translation tools part ways. Both are free to try in a browser. Both return a translation in seconds. What changes is what happens to your file, your tone and your language pair.

What actually happens when you hit translate

Neither tool looks up words one by one in a dictionary. Each one runs your sentence through a trained translation network (a neural machine translation model), which has compared millions of paired sentences in two languages and learned which wordings tend to travel together. When you paste a sentence, the system predicts the most likely equivalent in the target language, word by word, while keeping track of the whole sentence's meaning.

That shared idea explains why both tools handle idioms far better than the old dictionary-style translators did. It also explains why both can still slip: the network predicts what usually fits, not what your specific contract, joke or technical term actually requires.

Where DeepL tends to win

DeepL built its reputation on formal, business-flavored writing. If you paste a paragraph of a business email or a legal clause, the result often reads as though a careful human wrote it, with fewer awkward word choices. Its document translator accepts whole files — PDF, Word and PowerPoint — and rebuilds them with the original formatting, tables and layout largely intact. For a lease, a thesis chapter or a supplier agreement, that matters more than raw speed.

On the paid DeepL Pro plan, translated text is deleted right after processing and is not stored or used to train the network. If you handle client contracts or unpublished research, that policy is worth checking before you paste anything confidential.

Where Google Translate pulls ahead

Google Translate covers more than 130 languages, while DeepL handles roughly 35. That difference is not a detail — it decides which tool you can use at all. If you are traveling and need to read a menu, a train announcement or a street sign in a language DeepL does not support, Google Translate is the one that opens. Its camera mode translates signs live through your phone, and it is built into Android and the Chrome browser, so it is often already in front of you.

• Contracts, theses and business emails: DeepL, for its steadier formal tone and its ability to keep tables and layout in place.
• Travel, signs and less common languages: Google Translate, for its wider language list and live camera translation.
• Confidential material: check the privacy terms first — DeepL Pro deletes translated text after processing, while free tiers of either service may handle your text differently.

The mistakes both tools still make

A translation network predicts likely wording, so it can smooth over a term that should have stayed precise. Legal phrases, product names, units of measurement and names of people or places are common casualties. A clause that says "shall not be liable" can come back softer than the original, and a table of dates can shift by a day if the format is ambiguous.

The practical habit is simple: translate the document, then read the result against the original for anything that carries consequences — money, deadlines, obligations, medical instructions. If a sentence looks smoother than the source, that is often the sign to slow down, not to relax. For a signed contract, a qualified human translator is still the safer route.

So the next time a lease lands in your inbox, you already know the first move: DeepL for the document, Google Translate for the trip. Neither replaces your own eyes on the parts that matter.