Lost in translation: Where AI falls short on culture

Lost in translation: Where AI falls short on culture

Erika Gaitan is senior account executive at Vanguard Communication.

One area where AI still faces significant challenges is language, particularly when it comes to communicating effectively in non-English languages. As organizations increasingly use AI to translate and generate multilingual content, it’s important to understand the opportunities and the limitations of these technologies while being conscious of the dangers.

The promise of AI for multilingual communication

AI has significantly advanced multilingual content creation and translation. Today, AI tools can quickly translate documents, draft social media posts in multiple languages and help organizations reach diverse audiences more efficiently than ever.

For communicators, these advancements offer several key benefits:

Speed: Content can be translated or adapted in seconds, allowing teams to respond quickly and share timely information.

Scalability: Organizations can produce materials in multiple languages without significantly expanding staff or resources.

Accessibility: Information can reach communities that speak different languages, helping ensure more people can access important messages and resources.

For organizations working with diverse communities, whether in the public health, education or nonprofit sector, these capabilities can be especially valuable.

However, language is far more than words alone.

Language is cultural

Many AI models are trained primarily on English-language data. As a result, the tools may perform well in English but struggle with nuance, context or cultural meaning in other languages. Phrases that resonate in English may not translate well directly into Spanish, Mandarin, Arabic or other languages.

Literal translations can lead to:

  • Loss of cultural context
  • Incorrect tone or phrasing
  • Unintended meanings
  • Messages that feel unnatural or robotic

For example, the English phrase “mental health matters” might be translated literally into Spanish as “la salud mental importa.” While technically correct, some audiences may connect more with phrasing such as “cuidar nuestra salud mental es importante” (“taking care of our mental health is important”), which sounds more natural, conversational and culturally resonant.

Similarly, idioms often do not translate well. A phrase like “break the ice” could be translated literally into another language, where it might sound confusing or meaningless without cultural context.

In practice, this can result in communication that technically translates but doesn’t truly connect — therefore losing purpose and meaning.

The risk of ‘AI only’ translation

Organizations sometimes rely solely on AI translation tools to produce multilingual content quickly. While these tools are improving, they should not replace human review — particularly for sensitive or high-impact communications.

In these contexts, accuracy, tone and trust are essential. Even small translation errors can undermine credibility or cause confusion, especially during emergency situations. For example, a local government once translated “shelter in place” directly into Spanish as “refúgiese en el lugar.” While grammatically correct, the phrase can sound unclear, casual and not urgent. However, saying “permanezca dentro de su casa o edificio y no salga” ensures the message is immediately understood.

Where AI works best

Rather than replacing human translators, AI works best as a supporting tool. AI can draft initial translations or generate multilingual content ideas.

Human experts can then refine and adapt the content to ensure it reflects the cultural and linguistic realities of the intended audience.

Building more inclusive AI communication

As AI tools continue to evolve, organizations have an opportunity to build more inclusive communication strategies.

This means:

  • Prioritizing culturally appropriate translation
  • Working with native speakers and community experts
  • Testing AI-generated language before publication

Ultimately, effective multilingual communication requires more than technology; it hinges on an understanding of people, culture and context. AI can accelerate communication, but human insight remains essential for ensuring messages are meaningful, respectful and truly understood.

 

COMMENT

More From Author

Social media updates and new features to know this week

Social media updates and new features to know this week

 Why ‘we don’t know yet’ may be your best crisis response

 Why ‘we don’t know yet’ may be your best crisis response

Leave a Reply

Your email address will not be published. Required fields are marked *