Machine translation from €0.05/word Machine translation: fast, GDPR-compliant, from €0.05/word.
AI translation for companies that need to move volume into other languages quickly and securely. On EU servers, to ISO 27001, with documented post-editing to ISO 18587. Your data stays inside the protected environment, unlike with DeepL Free or ChatGPT.
Four ISO certificationsMore than 140,000 clients have trusted tolingo since 2007
Machine translation at tolingo at a glance
- What: AI translation of large text volumes, combined with post-editing by native specialist translators to ISO 18587.
- How: Our own AI engine on EU servers, ISO 27001-certified. Your data isn't used for AI training, NDA on request.
- Price: Light post-editing from €0.05/word, Full post-editing from €0.08/word. Repetitions are billed more cheaply through a translation memory.
- Difference: Unlike DeepL Free or ChatGPT, your text stays within the EU legal area and is reviewed by a specialist.
What is machine translation?
A definition that's practical for companies
Machine translation, MT for short, is the automatic translation of a text by an AI engine. Modern MT systems work with neural networks and reach a quality level for many language pairs that was unthinkable ten years ago. In practice, three points are decisive for companies: where the engine runs, what happens to the data, and who checks the result.
At tolingo, the MT runs on our own servers within the EU legal area, with ISO 27001 certification. A native specialist then checks the result in what's known as post-editing to ISO 18587. This second step is what sets the agency approach apart from free public-cloud tools. It catches the typical MT weaknesses: hallucinations, wrong specialist terminology, culturally unsuitable tone and inconsistencies across long documents.
When is machine translation worth it, and when not?
An honest rule of thumb
MT with post-editing is economically attractive as soon as three conditions come together: high volume, repetitive structures, and a medium to high tolerance for linguistic nuance. That suits product catalogues, internal knowledge bases, support articles, technical manuals or subtitles, for example. Here companies often pay half or a third compared with pure specialist translation to ISO 17100.
MT is less suitable for marketing texts with a brand voice, contracts with liability implications, high-end PR materials or texts in rare language pairs where the AI has too little training data. For English, German, French, Spanish or Italian the MT quality is very good; for Finnish to Hindi or Albanian to Vietnamese it's rather weak. A sound recommendation only comes after we've looked at your specific text, not as a blanket rule.
DeepL Free, DeepL Pro or tolingo MT?
What separates the three routes when it comes to company data
Many companies use DeepL Free because it's fast and free. For personal or confidential content that's a problem, because the data is subject to US servers. This overview shows the key differences.
| Criterion | DeepL Free | DeepL Pro | tolingo MT |
|---|---|---|---|
| Server location | DE/EU, but parent company in the US | EU servers, contractually secured | EU servers, own infrastructure |
| Data for AI training | Yes, possible under the terms | No, contractually excluded | No, ISO 27001 certified |
| Post-editing included | No | No | Yes, ISO 18587 |
| Translation memory | No | Glossary possible | Yes, grows across follow-up orders |
| NDA | No | Standard terms | Available individually |
| Price | free | from €7.49/month (tool) | from €0.05/word (incl. editing) |
For internal research snippets or non-critical keywords, DeepL Free is often enough. As soon as personal data, contracts, patient data, patent specifications or trade secrets are involved, the free option isn't a viable route. Here the requirements of GDPR Art. 28 and 32, trade-secret law and most corporate NDAs apply.
Light or Full post-editing?
Two quality levels, a deliberate choice
We don't deliver raw MT on its own. Every machine translation goes through post-editing to ISO 18587, in two levels depending on the intended use:
From €0.05/word. Technical accuracy and readability, without stylistic polish. Suitable for internal documents, training material, support articles.
From €0.08/word. Publication-ready quality, hard to tell apart from pure human translation. Suitable for customer communication, product descriptions, longer reports.
Exactly when Light is enough and when Full is needed is explained in detail on our editing and proofreading page. The full ISO 18587 workflow, with quality assurance and delivery formats, is described on the post-editing page.
Machine translation and the GDPR
Why DeepL Free is off-limits for many texts
When an employee pastes an employment contract, a patient file or an audit report into DeepL Free, that's a problem from a GDPR standpoint. The data leaves the protected environment, it's processed on servers whose parent company is subject to US law (CLOUD Act, FISA), and under the terms it can be used to improve the model. For personal data, GDPR Art. 28 (processing on behalf of a controller) and Art. 32 (technical and organisational measures) apply, and in many cases they rule out such a workflow.
tolingo MT runs on servers within the EU legal area, is ISO 27001-certified, and your data isn't used for training. For sensitive orders such as M&A contracts, compliance documents or annual reports, we set up dedicated project environments with restricted access on request. More on the general data-protection architecture under ISO 27001 at tolingo and on our page on GDPR-compliant translations.
What does machine translation cost?
Transparent word prices, predictable volumes
A worked example. A product catalogue of 25,000 words in Spanish, as Light post-editing: from about €1,250. As Full post-editing: from about €2,000. For comparison: the same translation as pure specialist translation to ISO 17100 comes in at from about €3,000. For recurring content such as follow-up releases of the same product line, the translation memory often cuts costs by 40 to 70 percent. For volumes from 100,000 words per year, we agree framework terms with tiered discounts.
You'll find the full price overview for all translation types on the pricing page. For express in 24 hours, a surcharge of 30 to 50 percent applies, depending on language pair and volume. Direct ordering of smaller volumes runs through the tolingo Shop.
Typical use cases
Where MT actually runs in companies
Thousands of product descriptions with a high repetition rate. MT with Full post-editing noticeably lowers the cost per language version.
Operating manuals and spare-parts catalogues with standardised terminology. Light post-editing is usually enough, see technical translation.
Support articles and knowledge bases with a high update frequency. The translation memory grows along with releases. More under website translation.
Training videos and conference recordings. We combine transcription with MT for scalable multi-language subtitling.
Frequently asked questions
What companies want to know before ordering
What's the difference between machine translation and pure specialist translation?
With MT, an AI engine translates the text, then a native specialist checks the result in post-editing to ISO 18587. With specialist translation to ISO 17100, a specialist translates directly and a second specialist revises. MT is cheaper and faster at large volumes; specialist translation is better for stylistically demanding texts and where liability matters.
Is machine translation GDPR-compliant?
At tolingo, yes. The AI engine runs on EU servers, is ISO 27001-certified, and your data isn't used for AI training. NDA on request. With free public-cloud tools like DeepL Free or ChatGPT, GDPR compliance for personal or confidential data is in many cases a problem, because the data is subject to US laws such as the CLOUD Act.
What does machine translation cost per word?
Light post-editing from €0.05/word, Full post-editing from €0.08/word. We don't offer raw MT without post-editing, because corporate clients almost always need a quality check. For recurring content, the translation memory reduces costs by 40 to 70 percent. The full overview is on the pricing page.
Which languages are supported?
Over 220 languages and language variants, in both directions. MT quality is very good for common language pairs (English, German, French, Spanish, Italian, Dutch, Portuguese). For rarer languages we often recommend Full post-editing straight away, or a pure specialist translation, because the AI models have less training data there.
What happens to my formatting?
The translated file comes back in the same format as the original: Word, PowerPoint, InDesign, XLIFF, JSON, HTML. Tags and layout are preserved. For complex layouts or languages with a different writing system (Arabic, Chinese, Japanese), we also offer layout adaptation.
How fast is delivery?
That depends on the post-editing level. The machine itself translates in seconds; the human review sets the pace. With Light post-editing, up to 10,000 words per working day is realistic; with Full post-editing, more like 4,000 to 5,000. A 25,000-word catalogue as Light is doable in three to four working days. With an express surcharge it's faster, depending on language pair and volume.
Can you connect a glossary or translation memory?
Yes. We maintain a terminology glossary and a translation memory per client. Both grow across orders and secure consistency over follow-up releases. With an API connection to your CMS or PIM, it runs in the background, more details under API integration.
Already have a translation?
Just have it post-edited, not re-translated.
If the text already comes from DeepL, ChatGPT or an internal tool and you just want it worked over, that goes directly through our editing & proofreading, with Light or Full post-editing to ISO 18587.
Get an assessment of MT for your next high-volume project.
Send us a sample file and briefly describe the purpose and language pair. We'll recommend Light or Full post-editing and give you a fixed price and delivery time. Also possible via the contact form.
Reply in ~10 minutes (within business hours) · ISO 18587 · ISO 27001 · EU serversGlossary
Terms around machine translation
- Machine Translation (MT)
- Automatic translation by an AI engine. Modern systems work with neural networks and capture the context of a text before translating.
- Post-editing (MTPE)
- Subsequent checking and correction of a machine translation by a specialist, governed by ISO 18587.
- ISO 18587
- International standard for post-editing of machine translations. Defines Light and Full post-editing as two quality levels.
- ISO 27001
- Standard for information security management systems. Governs technical and organisational measures to protect confidential data.
- Translation Memory
- Database in which already-translated segments are stored. On follow-up orders, repetitions are recognised automatically and reduce costs.