DeepL
DeepL is a German-built Language AI platform that translates text, documents, images and live speech across 100+ languages, plus a writing assistant and a REST API, with SOC 2 Type II and GDPR-grade controls.
What is DeepL?
DeepL is a Language AI platform built and operated by DeepL SE, a European company registered in Cologne, Germany, and founded there in 2017 by Jarek Kutylowski. It began as a translator and now presents itself as an end-to-end platform rather than a single tool, built on five products: Translator, Write, Voice, an API and a set of integrations, with two cross-cutting layers on top — Translation Flow for workflows and Customization Hub for brand control. What sets the engineering apart, according to the company, is that it runs a specialised in-house language model trained on proprietary data by thousands of language experts on its own supercomputing infrastructure, explicitly rather than calling public third-party models.
Text and document translation cover more than 100 languages; live meeting captions cover more than 40. Whole files can be dropped in and come back with their layout intact across PDF, Word, PowerPoint, Excel, HTML, images, InDesign and Contentful JSON. DeepL Write handles the other direction of the same problem — fixing grammar and syntax, rephrasing words or entire sentences, and shifting register between business, academic, enthusiastic or diplomatic tones. DeepL Voice translates speech in real time, either as live captions inside Microsoft Teams, Zoom and Google Meet, or on-device for face-to-face conversations, with a Voice API aimed at contact centres.
Distribution is deliberately wide: web, Windows and macOS applications, iOS and Android apps, Chrome, Edge and Firefox extensions, Word, PowerPoint and Outlook add-ins, Microsoft 365, Google Workspace, AWS Marketplace, and the CAT tools Trados Studio, memoQ and Across.
The commercial pitch leans as much on governance as on quality. DeepL publishes a SOC 2 Type II audit verdict and a Cure53 penetration test report as downloadable files, claims ISO 27001 and HIPAA alongside GDPR compliance, and offers bring-your-own-key encryption, SSO, multi-factor authentication, role-based permissions and audit logging. More than 200,000 business customers are claimed, with published cases including KBC Bank, Deutsche Bahn, JAE and Paysend.
What it does
- Translate text into more than 100 languages
- Translate whole documents while preserving the original layout
- Rewrite and polish text with adjustable style and tone
- Translate live speech in online meetings and face-to-face conversations
- Enforce brand terminology through glossaries, style guides and translation memory
- Translate images and screenshots
- Embed translation into your own products through a REST API
When to use DeepL / When not to
A quick filter to help you decide if DeepL is the right fit.
When to use DeepL
- Localisation and translation teams standardising terminology across dozens of markets
- Legal, pharmaceutical and financial teams needing accurate translation under strict compliance constraints
- Developers embedding translation, rewriting or real-time speech into their own products through the REST API
- Customer service and sales organisations running multilingual support without hiring per language
- Frontline and field staff who need face-to-face speech translation from a phone
When not to use DeepL
- Free-tier users who need to translate personal data, which the usage terms forbid outside a Pro subscription
- Buyers who require a published price before engaging, since no amount appears in the static pricing page
- Organisations that must contractually pin a specific hosting country, as DeepL keeps the right to choose where processing happens
- Anyone under 14, for whom the service is explicitly not intended
- Users in Belarus, Cuba, Iran, North Korea, Russia, Syria or the occupied regions of Ukraine, where access may be restricted
How to use DeepL
A typical end-to-end flow, from setup to results.
- Open the free web translator and paste up to 5,000 characters, with no account required
- Create a DeepL account with an email address and password to keep glossaries across devices
- Drag a PDF, Word, PowerPoint, Excel or InDesign file onto the translator to translate it in place
- Click any word in the translation to see alternatives, and watch the surrounding sentence adjust itself
- Save a chosen alternative into a glossary so that it is reused in every later translation
- Switch to DeepL Write, use Rephrase word or Rephrase sentence, then pick a writing style and a tone
- Install the desktop app, a browser extension or the Word, PowerPoint and Outlook add-ins to work in place
- Start a Pro trial from the pricing page to unlock unlimited text, larger files and CAT tool integration
- Request a free API key, then call POST /v2/translate with your text and the source and target languages
- For a company rollout, configure SSO, multi-factor authentication, role-based permissions and domain management
Pros & Cons
Pros
- Translation quality backed by material the company publishes rather than merely asserts
- A European publisher under GDPR, with personal data stored on servers inside the EEA
- Certifications that can actually be checked: the SOC 2 Type II audit verdict and a Cure53 penetration test report are downloadable files
- On paid plans, submitted texts are deleted once processed and are not used for training
- A permanent free tier alongside a free trial on the paid plans
- Unusually broad delivery: web, Windows, macOS, iOS, Android, browser extensions, Office add-ins and API
- Layout survives translation, including on PDF and InDesign files
Cons
- No price is readable without JavaScript, as the pricing page ships only placeholders
- On the free tier your content trains the models, and translating personal data there is contractually forbidden
- No documented opt-out switch: excluding your data from training means moving to a paid plan
- No specific hosting country is committed to, and the terms let DeepL decide where processing happens
- Governing law and exclusive jurisdiction are England and Wales, despite a German publisher
- The help centre is unreachable outside a browser, which slows down due diligence
- Voice-to-voice inside online meetings is still announced as coming soon
Pricing & Plans
DeepL operates on a freemium basis. A permanent free plan is available: the web translator can be used without an account, capped at 5,000 characters per translation, and a free DeepL API key is offered to developers. Paid Pro subscriptions are sold on top, preceded by a free trial during which nothing is charged in the first month and which converts automatically into a paid subscription at its term unless cancelled beforehand. No entry-level amount can be quoted here: the pricing page renders its figures client-side and publishes no amount in its static markup, so neither the lowest price point nor its currency could be established from the site. API usage is billed by characters for text and by audio minutes for speech.
- Starter — entry-level Pro plan
- with file uploads capped at 10 MB
- Advanced — all Starter features plus CAT tool integration and file uploads up to 20 MB
- Ultimate — all Advanced features plus the highest limits
- Advanced + Write Pro — Advanced bundled with the Pro writing assistant
- Ultimate + Write Pro — Ultimate bundled with the Pro writing assistant
- Voice Starter — entry-level plan for real-time speech translation
- Business — enterprise plan carrying a 99.0% annual uptime commitment instead of 97%
- DeepL API Free — free developer key
- without CAT tool integration or web translator access
- DeepL API Developer — paid developer tier
- DeepL API Growth — production API tier
- designed for at least 600 characters per minute per customer
- DeepL API Pro — API tier covered by the Pro data handling commitments
- Enterprise — quoted through the sales team
- with no public price
Data, GDPR & hosting
A consolidated view of how DeepL handles your data.
GDPR overview
GDPR compliance is documented concretely rather than claimed in passing. The controller is DeepL SE, Maarweg 165, 50825 Cologne, and the policy names both the GDPR and the German BDSG as the applicable framework. An external Data Protection Officer is designated: Dr. Christian Lenz of dhpg IT-Services GmbH, reachable at datenschutz(at)dhpg.de. Data subject requests go to privacy(at)deepl.com, and an Article 28 data processing agreement is available through sales(at)deepl.com. No Article 27 representative exists, and none is required, the controller being established inside the EU. Legal bases are stated case by case: Article 6(1)(b) for Pro content, 6(1)(f) for technical logs, 6(1)(a) for user interviews. The supervisory authority named is the LDI Nordrhein-Westfalen in Düsseldorf. The policy was last updated in June 2026.
Who owns the data?
Under the Pro terms, every right in the content you submit, in the processed output and in your Customer Training Data stays with you. DeepL takes only a non-exclusive worldwide licence to temporarily store, modify, process, translate, improve and transmit that material in order to deliver the service, and may sub-license those rights to its subcontractors strictly as far as delivery requires. Customer Training Data such as glossaries is used only for that customer's own translations. The free tier differs: uploaded content and its translations are processed for a limited period to train and improve DeepL's neural networks. For Voice, DeepL acts as a processor under a data processing agreement signed with the customer.
Reuse rights
Users keep ownership of what they submit and of the translations returned, so nothing in the terms requires asking DeepL's permission before reusing that output, commercially or otherwise. What differs is what DeepL may do in return. On paid Translator Pro, API Pro and Write Pro plans, submitted texts and documents are held only as long as producing and transmitting the result requires, then deleted, and the enterprise security page states that texts are never stored or used for model training without consent. On the free translator the trade-off is explicit and reversed: content and its translations, including the corrections users make to suggestions, feed model training for a limited period, and translating personal data there is contractually prohibited.
Data retention & training
Hosting summary
DeepL states that personal data may be stored on its own servers located within the European Economic Area, and the company is established in Germany, which places the arrangement squarely under EU law. That commitment is regional rather than national: no specific hosting country is named, and the Pro terms explicitly reserve DeepL's right to determine where processing takes place, including on third-party cloud infrastructure. Two exceptions are documented by name. Voice for Meetings relies on Microsoft Ireland Operations Limited and Amazon Web Services EMEA in Luxembourg purely to host the meeting bot and relay audio: neither transcription nor translation happens there, and no data is stored on those servers. The developer documentation site is hosted by GitBook, with Fathom analytics processed in the United States. Cloudflare fronts the website as a CDN but sees only metadata, traffic between the device and DeepL's servers being encrypted. Salesforce CRM data is kept on servers in the EU, while payments, support ticketing and marketing email rely on processors outside the EU under data processing agreements.
Where DeepL works
Country-level availability.
Not available in
Things to keep in mind
Risks and trade-offs to weigh before adopting DeepL.
- Leaning on machine translation for high-stakes legal or medical wording can erode the habit of having a qualified human check it
- The convenience of the free tier invites people to paste customer or employee data that the usage terms forbid there
- Fluent output reads as authoritative, and errors in a language you do not speak are almost impossible to catch yourself
- Regular use of a writing assistant can gradually flatten a personal or a brand voice
- Relying on translation rather than language skills can quietly reduce a team's own multilingual capability
- The free trial converts automatically into a paid subscription at its term, which can produce an unintended charge
- Payments, support ticketing, analytics and marketing rely on processors outside the EU, so some personal data does reach the United States
Setup & Integrations
Technical difficulty
Low for everyday use, moderate for a company rollout. The web translator works immediately with no account and no installation, and the desktop apps, browser extensions and Office or Google Workspace add-ins install in a click. The API is equally undemanding: request a free key, then send a REST call with the text and the source and target languages, with sample code provided. Customization Hub is presented as a matter of minutes without heavy technical enablement. Real effort begins at company scale, where SSO, multi-factor authentication, role-based permissions, domain management and encryption keys must be configured.
Deployment
Apps stores
Integrations
Supported languages
Behind DeepL
Fundraising
Social
Resources
All the official URLs gathered for verification and reference.
Alternatives
Tools that compete with or complement DeepL.
Frequently asked questions
Is DeepL free to use?
How many languages does DeepL support?
Are my texts used to train DeepL's models?
Can I translate personal data with DeepL?
Does DeepL offer an API?
Which security certifications does DeepL hold?
Where is my data hosted?
Is DeepL available everywhere?
Who is behind DeepL?
How do I cancel a subscription?
Should you pick DeepL?
DeepL is a mature, actively developed product backed by an identifiable European company with substantial funding behind it. Its positioning is unusually coherent: translation quality on one side, and compliance and security credentials on the other, each supported by material you can actually inspect rather than by claims alone — a downloadable SOC 2 Type II audit verdict, a published penetration test report, a Trust Center and a status page. For an organisation that has to justify a language tool to its own security or legal function, that combination is rare and it is the strongest argument here.
The product surface is genuinely broad. The same platform covers a free web translator with no account, layout-preserving document translation, a writing assistant, real-time speech translation in meetings and in person, and a REST API — which means a single vendor decision can serve an individual, a localisation team and an engineering team at once.
Two reservations deserve to be stated plainly. The first is commercial transparency: the pricing page renders its figures client-side and its static markup carries no amount at all, so no entry price could be established from the site itself, and the plan names are known while their cost is not. The second is the free tier. Content submitted there is used for a limited period to train and improve the models, and the usage terms explicitly forbid putting personal data into it — a boundary that is easy to cross without noticing. Paying moves you to a much stricter regime, where texts are deleted after processing and are not used for training.
Verify the current prices in a browser before committing, and match the plan to the sensitivity of what you intend to paste into it.
- Choosing a selection results in a full page refresh.
- Opens in a new window.