How AI Translation Could Misrepresent Your Brand Message

Once on a family holiday to New York, we took a tour of the United Nations. I remember a cathedral-like atrium and the sense that important things were being done, but what really stayed with me was the UN interpreters’ room. 

It had a filmic, ’80s sci-fi quality — a sound insulated booth with a panoramic view of the cavernous assembly hall. Sat inside was a group of highly focused professionals arranged in descending rows that reminded me of NASA mission-control. A speech was underway, and they were all speaking into small microphones attached to diode-lit dashboards. We watched quietly while delegates on the floor listened to their designated interpreter’s live translation via headset.

It was very high-tech for its time, but the essential operation was all human. 

Then and now, translation meant more than just converting words and phrases from one language to another, it required deciphering, contextualising and reading between the lines. What the UN interpreters delivered wasn’t so much a live transcript as a cultural and linguistic localisation. 

This localisation put the delegates listening to the speech in a much better position to understand the nuances, allusions, metaphors and double entendres that might have been lost in a mere transposition of words. They gained a fullness of meaning that allowed them to better guide their government’s diplomatic responses. 

Brands, with thousands, millions — even billions of pounds on the line, deserve to have their communications treated with the same respect. A chorus of tech enthusiasts is now telling us that they can provide accurate, high-quality translations quickly and easily, thanks to the arrival of AI. 

Translation by machine isn’t new. Who can forget the early iterations of Google Translate and its endless list of fails? Today we have Deepl, ChatGPT, Wordly, the newer/smarter version of Google Translate, and a host of others. They all claim to use neural networks, machine learning and large language models (LLMs) to replicate what a professional translator takes years to master. 

There’s no doubt that they’re better than what came before, but should marketers really rely on them? Can advanced software accurately deliver a brand’s intended meanings or emotional responses from one cultural context to another? 

The short answer is no, at least for now — and here’s why. 

Whether it’s textual or verbal, communication between two parties always involves more than factual statements and logical responses. Humans infuse their interactions with sarcasm, irony, pathos and other devices that add hidden layers of meaning a machine might miss.  

Sometimes these happen by simply changing the emphasis on a word or juxtaposing an image in a certain way against another. In a media-obsessed world, people also speak with an awareness of the current zeitgeist and its catch phrases and pop-culture references.  

Maybe an AI will recognise a popular internet meme, but will it understand how to reference it in a sales nurture — or know if it’s even appropriate in a different cultural setting? 

Worried your key messages are being lost in translation? Ask about Purecontent’s localisation service. It’s more than word-to-word conversion. We’ll ensure your quality, tone of voice and meaning are all intact. Contact us today. 

When you’re trying to promote a message, product, service or ESG commitment across linguistically diverse groups, misunderstandings matter. The consequences can be dire, from lost revenue to loss of reputation. At minimum, the translation of text from one language to another needs to result in an output that is factually correct and meaning-specific. 

With AI-driven translation, this remains a serious worry. An assessment of Google Translate for emergency department instructions conducted in 2021 found that accurate meaning was retained 82.5 percent of the time. In a healthcare setting that leaves a lot of room for error, but dig deeper and the variation is more extreme. Translations to Spanish were accurate more than 90 percent of the time, but translations to Armenian got it wrong 45 percent of the time. 

Bear in mind that these were clinical instructions written for a science-literate audience. What would the results have been if more nuanced and potentially ambiguous marketing messages had been tested the same way? How accurately would ChatGPT render the emotional content and mental associations attached to your brand and convey them in Farsi, Mandarin, Estonian, or another of the world’s estimated 7,000 languages? 

In marketing, a bad translation can have impacts that go well beyond inconvenience or frustration. 

• Lost Sales: If a customer or prospect receives a poorly translated communication, they might also perceive a lack of professionalism that extends to other areas of your business, potentially even taking their business elsewhere. 
• Delays: Re-doing a bad translation takes time and might need input from numerous vendors and internal stakeholders if the errors are bad enough. You could get caught in a new round of approvals, resulting in missed deadlines and revised plans.
• Damaged Reputation: Sometimes a word or phrase that’s harmless in your native language can have different or even negative connotations in the target language. Customers might be confused, or even insulted. 
• Increased Costs: Bad translations need to be re-done, adding time and cost — especially if materials or displays have already been printed and assembled. 

A programmer might tell you that every activity can be broken down into an algorithm, but great translation draws on the human power of synthesis — a unique set of mental operations that machines simply can’t perform.  

How it works is complex and mysterious. We still don’t fully understand how the brain can ingest a weighted mix of linguistic, textual, verbal, cultural, social, technical and media inputs from one language and then generate an accurate output in another.  

It’s part of the magic of being human. No collection of code will ever entirely replace it. 

AI has utility for translating simple text or giving you a flavour of what’s being said in another language, but only human translation can give you the confidence that your messages and values are being delivered accurately and in the right context. 

What we’re talking about now is quality. Of course, there are levels of quality on the human side to consider too. Do you want a UN-level professional, with years of education and experience perfecting their craft, or someone on a gig-economy platform doing translation as a side hustle? 

Tired of trusting translation to non-professionals? Ask us about content localisation. We’ll ensure your quality, tone of voice and meaning are all intact. Contact us today. 

Mark de Wolf is an award-winning copywriter and freelance journalist specialising in AI-free, human enriched content. Born in Toronto. Made in London. Based in Zürich. Reach him at markdewolf.com or on Linkedin at https://www.linkedin.com/in/techcopy/.