When AI misses the meaning: why human expertise remains essential in life science translations

Nov 13, 2025

Artificial intelligence (AI) has become a valuable tool in translation. Large Language Models (LLMs) can process millions of words in seconds, reducing costs and accelerating workflows. Yet, as recent findings show, even the most advanced AI systems still struggle with what makes language truly human: context, idioms, and culture.

A study published by Slator revealed that LLMs often fail to accurately translate idiomatic expressions and culturally specific language, producing literal, sometimes misleading results. While this might seem like a minor issue in everyday communication, in life sciences the consequences can be far-reaching, impacting patient understanding, regulatory compliance, and brand credibility.

When AI translates the words but misses the meaning

Idioms, metaphors, and cultural references are deeply embedded in language. When an AI model translates the phrase “a bitter pill to swallow” literally, for example, it may lose the intended meaning of something difficult to accept. According to Slator’s report, these kinds of errors remain common across major LLMs, even in high-resource languages like French and German.

In life sciences, this is not just a linguistic problem. It is a human problem. Patient-facing materials, such as leaflets, informed consent forms, or health app instructions, must be written in clear, relatable language that resonates with the target audience’s culture and communication style. A literal or culturally tone-deaf translation can confuse patients or even discourage them from following correctly the instructions.

Culture and clarity in patient communication

Cultural adaptation goes beyond words. It is about how information feels to the reader.
In one country, a direct tone might convey clarity and confidence. In another, it could sound abrupt or insensitive. Similarly, color symbolism, imagery, and examples used in patient education materials often need localisation to avoid misinterpretation.

This cultural layer is precisely where human expertise becomes indispensable. A trained medical linguist understands both the medical terminology and the patient’s perspective. They can judge when a phrase needs rewording for cultural resonance or patient empathy, something AI alone cannot reliably do.

When it comes to patient safety and understanding, “close enough” is not enough. Every word carries weight.

Why “good enough” translations fall short in regulated industries

Life science companies operate in one of the most regulated linguistic environments in the world. From clinical trial documentation to device labeling, every translation must align with both scientific accuracy and regulatory precision. A literal translation that fails to convey the correct context or dosage instruction could trigger non-compliance or even put patients at risk.

Regulators expect linguistic consistency across all markets. They also expect companies to demonstrate traceability, knowing who translated what, when, and how quality was verified. AI output, if not reviewed by qualified experts, does not meet those standards.

That is why, at Novalins, AI is never a replacement. It is a tool within a certified, human-centered process.

AI as an accelerator, not an authority

AI undoubtedly brings measurable benefits: faster turnaround times, lower costs, and the ability to pre-translate repetitive or low-risk content. It can analyse previous translations, identify patterns, and support terminology management across large document sets.

However, AI’s output must always be validated by life science professionals who understand the subject matter. Human reviewers ensure that translations are not only correct but also compliant, consistent, and culturally appropriate. In other words, AI can help translators work smarter but it cannot replace their judgment or expertise.

At Novalins, we integrate AI within our ISO-certified workflows. This means every AI-assisted translation passes through rigorous quality steps, including expert medical review and regulatory validation. The result is faster delivery and reduced cost, without compromising patient safety or linguistic precision.

Transparency builds trust

One of the growing concerns in the industry is how AI is used. Transparency is key. Life science companies must know when AI contributes to the translation process and to what extent. At Novalins, our clients are always informed whether AI is applied for pre-translation, terminology optimization, or other stages of the workflow.

By combining transparency with human expertise, we ensure that AI remains a supporting actor, not the lead.

Conclusion: technology enhances, humans ensure

The Slator findings remind us that even the most sophisticated AI systems still struggle with meaning, emotion, and cultural nuance. In the life sciences, these are not minor details. They are central to effective communication and compliance.

AI continues to reshape how multilingual content is managed, but human expertise remains irreplaceable. Medical translators, reviewers, and cultural consultants ensure that every word is not only accurate but meaningful to its audience.

Interested in seeing how AI and expert review can optimize your multilingual projects?

Try a free pilot project with Novalins’ AI-assisted medical translation workflow and experience how technology and human expertise can work together seamlessly.