DeepL

DeepL Translator is a text and document translation engine. You paste text, drop in a file, or call the API, and it returns the same content in another language. It is not a chatbot, it does not answer questions about the text, and it does not rewrite your English to sound better — that is DeepL Write, a separate product from the same company with its own subscription and a much shorter language list. Confusing the two is the most common mistake in write-ups of DeepL, and it matters commercially: DeepL's own help centre states plainly that access to paid DeepL Write is not included in a DeepL Pro plan.

For most of its life DeepL's pitch was "fewer languages, better ones." That has changed. The current help-centre table lists well past 100 languages, including a long tail that DeepL never used to touch — Acehnese, Cebuano, Kapampangan, Pangasinan, Quechua, Sicilian, Tswana, Wolof, Xhosa, Zulu, plus Latin and Esperanto. It also handles variety selection with more care than most engines: English targets split into American and British when the source is one of about eighteen languages, Portuguese splits Brazilian from everything else, Spanish offers European and Latin American targets (the latter only on DeepL's next-generation model), Chinese offers simplified and traditional targets, French offers France and Canadian, German offers Germany and Swiss, and Arabic is Modern Standard.

The caveat is in the same document: not every listed language is supported for glossaries, formal/informal tone, or alternative phrasings. If those features are why you are paying, check the per-feature language lists before you commit a workflow to a pair.

DeepL publishes its own blind-preference comparisons against other engines on its quality page. Those are vendor-run studies, and they should be read as such — a company measuring itself. What is harder to dismiss is the consistency of the reception: professional translators, localisation teams and European businesses have gravitated to DeepL for the better part of a decade specifically for how the output reads, not for how many boxes it ticks. The mechanism is plausible, too. DeepL trained heavily on parallel European corpora, and its output tends to restructure sentences into natural target-language word order rather than tracking the source clause by clause.

That advantage is narrower than the marketing implies. On a low-resource pair from the newly added list, there is no reason to assume DeepL is ahead of anyone — and no published evidence that it is.

Text translation is a commodity; keeping a 40-page PowerPoint intact through a language change is not. DeepL's document translation covers Word, PowerPoint, Excel and PDF, retaining layout, and it is metered: free accounts get a small monthly allowance, paid tiers get more, and the higher tiers make the translated file fully editable rather than locked. For most teams this single feature, not raw translation quality, is what justifies the line item.

Two features separate DeepL from consumer translation. Glossaries pin terminology — brand names, product names, internal jargon — so the engine stops "helpfully" translating them. Formality control lets you force a formal or informal register, which matters enormously in German, French, Japanese and Spanish and not at all in English. Both are tier-gated and, again, language-gated.

Around them sits the integration surface: desktop apps for Windows and macOS, iOS and Android apps, Chrome, Firefox and Edge extensions, add-ins for Word, Outlook, PowerPoint and Google Workspace, CAT-tool support for professional translators, and an API with a free key and a paid per-character tier. DeepL has also pushed into voice with a real-time translation product and into agent tooling, which tells you where the company thinks the growth is.

DeepL sells a free tier, a set of per-user Pro tiers (Starter, Advanced and Ultimate), a separate Write Pro subscription that can be bundled with Pro, and API plans with a free and a paid level. Prices are per seat and vary by country, and the pricing page renders entirely in JavaScript — as of August 2026 no figures appear in the page source, and third-party trackers disagree with each other by several dollars a month. The structural point stands regardless of the number: DeepL charges by headcount, not by volume, so a small team pushing enormous quantities of text pays little, while a large team translating occasionally pays a lot. If your usage looks like the second case, price the API instead.

Google Translate covers 249 languages and dialects, is free at consumer scale, and now runs Gemini models with live speech translation built in. Microsoft's translation lives mostly in Azure and inside Office, Teams and Edge, where it is bought as infrastructure rather than chosen as a product. DeepL competes with neither on breadth. It competes on the sentence — on whether the paragraph you hand a client reads like it was written for them.

Take DeepL if you move European-language business documents, need glossaries and formality control, and care that the output does not read like a machine. Skip it if your pair sits in the long tail, if you need the widest possible coverage, or if what you actually want is a writing assistant — in which case the product you are looking for is DeepL Write, sold separately.