You’ve been there. You finish your novel, your memoir, your nonfiction guide, and you start thinking about international readers. Millions of people who might love your book if only they could read it. So you paste a paragraph into Google Translate, hit enter, and get back something that reads like it was assembled by a committee of dictionaries. The words are technically correct. The meaning is roughly intact. But the voice, the thing that makes your writing yours, is gone.
That gap between “technically correct” and “actually sounds like a human wrote it” is the entire reason DeepL exists.
A Dictionary That Became Something Bigger
DeepL’s origin story doesn’t start with a translator. It starts with a dictionary.
In 2009, a former Google researcher named Gereon Frahling founded Linguee in Cologne, Germany. Linguee wasn’t a traditional dictionary. Instead of matching words to definitions, it crawled the web for professionally translated documents and showed how phrases were actually used in real translated text. Need to know how a native German speaker would express “it dawned on me”? Linguee found examples from professional human translations and showed you the context.
Over the next several years, Linguee quietly amassed one of the largest collections of high-quality bilingual text on the internet. Millions of sentence pairs, all sourced from professional translators and verified for accuracy.
In 2012, a computer scientist named Jaroslaw Kutylowski joined as CTO. Kutylowski (who goes by Jarek) grew up switching between Polish and German, started coding at ten, and holds a PhD in computer science with a focus on mathematics. He began experimenting with neural networks to improve Linguee’s translation verification. The team noticed something unexpected: the neural networks, trained on Linguee’s decade of curated bilingual data, could do full machine translation, and do it significantly better than existing tools.
They had spent years building what turned out to be a uniquely powerful training dataset for a translator, without originally intending to build one.
In August 2017, they launched DeepL Translator. Kutylowski became CEO in 2019, and the company has since grown to over 900 employees, serving more than 100,000 businesses in 63 markets, with a valuation near $2 billion.
When asked about competing with tech giants, Kutylowski is matter-of-fact: “Unlike the US tech giants, we focused on our one goal from the very beginning. For Google, the translator is just a small sub-project, whereas at DeepL we are all pulling in the same direction at full steam.”
What DeepL Actually Does for Authors
At its core, DeepL translates text between 36 languages. You paste in your words (or upload a document), choose a target language, and get a translation back. So far, nothing you haven’t seen before.
The difference is in what comes out the other side.
DeepL’s translations read like a person wrote them. Not a particularly creative person, not someone who’s going to preserve your metaphors perfectly, but a competent bilingual person who understands context, syntax, and idiom. Where Google Translate might render a colloquial English phrase into awkwardly literal German, DeepL tends to find the natural equivalent.
For practical author workflows, DeepL handles three things well:
Text translation. Paste up to 50,000 characters per month on the free plan (roughly 8,000 to 10,000 words) and translate between any of the 36 supported languages. Paid plans increase this to 300,000 characters on the Individual tier, a million on Team, and unlimited on Business.
Document translation. Upload Word documents, PDFs, PowerPoints, or text files and get translated versions back with formatting preserved. This matters if you’re translating a formatted manuscript or a media kit for foreign publishers. The free tier allows one file per month; paid plans scale up.
A glossary system. You can create custom glossaries that tell DeepL how to handle specific terms. If your fantasy novel uses “Shadowmere” as a proper noun that shouldn’t be translated, or if you want “the Guild” consistently rendered as “die Gilde” in German, glossaries enforce that. Business plans offer up to 250 glossaries with unlimited entries.
DeepL also offers a writing assistant called DeepL Write, which handles grammar, style, and tone adjustments. It’s a secondary feature, not the main draw, and currently works in a limited set of languages.
The Thing That Makes It Different
Every translation app uses AI. The question is what that AI was trained on.
Most machine translation systems learn from the open internet. They consume billions of pages of text, learn statistical patterns, and produce translations that reflect the average quality of what’s out there. This works well enough for getting the gist of a restaurant menu or a news article.
DeepL’s AI grew up differently. It learned from Linguee’s curated corpus of professional human translations, over seven years of bilingual text pairs that were verified for accuracy before they ever entered the training data. The difference is the equivalent of learning a language by reading published literature versus learning it from comment sections.
In blind tests, professional translators found that DeepL’s output required two to three times fewer edits than translations from Google Translate or ChatGPT. For authors, that’s the gap between “I can send this to my foreign publisher with some light editing” and “I need to hire someone to redo this from scratch.”
DeepL is also notably private. The company states that it does not use subscriber text to train its models. For authors translating unpublished manuscripts, this is a real consideration. Your unfinished novel isn’t becoming training data for someone else’s AI.
Where DeepL Fits (and Where It Doesn’t)
DeepL is excellent at producing a clean first-draft translation that a human can refine. It’s the best machine translation tool available for European language pairs, and it’s strong with Japanese, Korean, and Chinese as well. If you write in English and want to explore whether your book could find an audience in Germany, France, Spain, Portugal, or a dozen other markets, DeepL gives you the fastest path to a readable translation.
But let’s be honest about the limits.
36 languages, not 130. Google Translate covers three times as many languages. If you need Hindi, Swahili, Tagalog, or Thai, DeepL can’t help you. Its strength is depth over breadth, particularly in European languages.
Machine translation is not literary translation. DeepL produces functional, natural-sounding prose. It does not produce art. Wordplay, cultural references, poetic rhythm, narrative voice: these are the things that make translated literature sing, and they require a human translator who understands both the source and target cultures. If you’re publishing a translated edition of your novel, DeepL is your starting point, not your finished product.
The free tier runs out fast. 50,000 characters sounds like a lot until you realize that’s roughly one or two chapters. Authors translating full manuscripts will need a paid plan, and even the Individual tier’s 300,000 characters per month (about 50,000 words) covers only a short book.
It’s a translator, not a translation management system. If you’re working with a team of translators on a large project, DeepL lacks the workflow features (segment history, real-time collaboration, advanced QA checks) that full translation management platforms provide.
Pricing at a Glance
DeepL’s free tier is genuinely useful for testing and short projects. Paid plans are billed monthly, with discounts available for annual billing:
The Free plan includes 50,000 characters and one file translation per month. The Individual plan at $10.49/month bumps that to 300,000 characters and three file translations. The Team plan runs $34.49 per user per month with a million characters and 20 file translations. And the Business plan at $68.99 per user per month removes the character cap entirely and allows 100 file translations.
All paid plans include a 30-day free trial. Annual billing brings prices down (the Individual plan drops to $8.74/month, for example). There’s also a Developer API with a free tier at 500,000 characters per month and a Pro tier starting at $5.49/month plus usage-based charges.
The Bottom Line
DeepL exists because one team in Cologne decided that machine translation should actually sound like a person talking. They spent a decade building the dataset to back that up, and the results show. For authors exploring international markets, translating correspondence with foreign publishers, or simply wanting to understand how their work reads in another language, DeepL is the best general-purpose translation tool available right now.
It won’t replace a skilled literary translator for your published edition. But it will give you a first draft good enough that your translator spends their time polishing instead of rewriting. For a lot of authors, that changes the math on whether reaching international readers is worth pursuing at all.
