What medical machine translation really involves, and why it’s never just copy paste 

Machine translation is often perceived as an instant solution, fast, automated, and inexpensive. 
In medical and regulatory contexts, this perception couldn’t be further from reality. 

When machine translation is used responsibly for medical content, it becomes a highly engineered process, involving expert decision-making, multiple technologies, rigorous testing, and strict quality and compliance controls. 

Here is what actually happens behind the scenes. 

Step one, selecting the right technology, not just one engine 

There is no such thing as a single machine translation engine that works well for all medical content, all languages, and all document types. 

For each project, we evaluate: 

• language pair and target market 

• medical domain and risk level 

• regulatory purpose of the content 

• required terminology consistency 

• downstream use (submission, labeling, software UI, IFU, etc.) 

This is why we work with several machine translation and language AI technologies, rather than relying on a single solution. 
Each tool is tested and benchmarked on real medical samples, not generic text. 

Importantly, we only use technologies that: 

• comply with ISO 27001 information security standards 

• operate under premium licences 

• do not use client data to train models 

Data protection and confidentiality are not optional steps, they are foundational. 

Step two, testing and validation by medical language experts 

Before any tool is integrated into a production workflow, it is: 

• tested on representative medical content 

• reviewed by subject-matter experts 

• evaluated against predefined quality criteria 

This step allows us to identify: 

• typical error patterns 

• terminology weaknesses 

• language-specific risks 

• suitability for regulatory use 

Only tools that pass this validation phase are used, and even then, only within clearly defined scopes. 

Step three, connecting technology to a controlled TMS environment 

Once the most performant solution is selected, it is not used in isolation. 

The technology is integrated into a structured translation management system (Phrase) to ensure: 

• traceability 

• version control 

• audit readiness 

• controlled workflows 

At this stage, we also: 

• upload and align client translation memories 

• integrate approved glossaries and terminology rules 

• configure QA checks specific to the client and product 

This is what ensures consistency with legacy content and regulatory expectations. 

Step four, pre-DTP preparation to avoid critical errors later 

Medical translation does not start with the text alone. 

Before translation, files are prepared to: 

• preserve layout integrity 

• protect non-translatable elements 

• avoid segmentation issues 

• ensure compatibility with DTP workflows 

Skipping this step often leads to formatting issues, truncated text, or compliance risks that only appear at submission stage. 

Step five, human medical review in line with ISO 18587 

When machine translation is used for medical content, post-editing is not a generic review step. 

Our workflows are designed in line with ISO 18587, the international standard for post-editing of machine translation output. This means that: 

• post-editing is performed by qualified linguists with medical expertise 

• reviewers are trained specifically for machine translation post-editing, not just traditional translation 

• quality criteria are clearly defined and documented 

• the level of intervention is adapted to the intended use and regulatory risk of the content 

ISO 18587 ensures that machine translation output is: 

• accurate and complete 

• appropriate for regulatory and clinical contexts 

• consistent with approved terminology and style 

• safe for downstream use 

In other words, post-editing is treated as a controlled, auditable process, not as a quick correction pass. 

Why ISO standards matter in machine translation workflows 

Between ISO 27001 for information security and ISO 18587 for post-editing quality, machine translation workflows can be aligned with the same expectations that apply to traditional medical translation. 

This standards-based approach provides: 

• traceability for audits 

• clarity on roles and responsibilities 

• reassurance for regulatory and QA teams 

• a defensible quality framework when machine translation is used 

Machine translation does not replace quality standards. 
It must operate within them. 

Step six, post-DTP checks and final quality controls 

After layout and formatting: 

• final linguistic checks are performed 

• content is reviewed in context 

• QA checks are rerun to ensure no issues were introduced during DTP 

This step is essential for labels, IFUs, packaging, and software interfaces, where layout issues can directly affect compliance. 

Conclusion, machine translation is a system, not a shortcut 

In regulated environments, machine translation is not a faster alternative to traditional translation. It is a standards-based system that only delivers value when designed, validated, and monitored correctly. 

What clients receive is not “machine translation”, but a controlled workflow, built around: 

• carefully selected and validated technologies 

• secure, ISO 27001–compliant environments with premium licences 

• structured TMS workflows with full traceability 

• qualified medical post-editors working in line with ISO 18587 

• pre-DTP and post-DTP controls to ensure regulatory readiness 

This is the difference between using machine translation as a risk, or using it as a reliable, auditable component of a medical localization strategy. 

Want to understand how this applies to your content? 

Every medical product, language pair, and regulatory pathway is different. 

If you’d like to understand how machine translation could be safely applied to your specific use case, we’re happy to walk you through our workflows in detail and discuss validation options. 

Because in medical translation, the work you don’t see is often the most important.