Why AI Translation Still Falls Short — And What We Learned From a One-Page Test

We recently ran an experiment. 

Curious about the capabilities of a specialized AI translation tool, we translated a single page of English content into French and asked one of our most trusted human translators to review the output. 

The verdict? 

It wasn’t pretty. 

What we hoped would be a clean, if imperfect, first draft turned out to be riddled with confusing phrases, misused terms, awkward syntax, and outright errors. Some lines were so jumbled that our translator labeled them "incomprehensible." 

Here’s just a sampling of her feedback: 

• Incomprehensible structure: “This sentence doesn’t make ANY sense in French, especially the second part. We would never translate the English sentence that way.” 

• Wrong tone and word choice: “‘Nous ajoutons de la valeur’ — YIKES. That’s not how we say ‘we add value’ in French. Very poor quality!” 

• Terminology errors: “‘Fabricants automobiles’ is incorrect. It should be ‘constructeurs automobiles.’” 

• Clunky literalism: “‘Maintenir’ isn’t wrong, but it’s not good either. A skilled translator would use ‘mettre en place’ in this context.” 

• Grammar flubs: “Should be ‘preuve d’un,’ not ‘preuves pour un.’ Wrong tense, wrong structure.” 

• Wrong verb meanings: “Using ‘impliquer’ for ‘involve’ is misleading —  it implies getting caught in something unpleasant. In this case, ‘participer’ is correct.” 

• Missing critical words: “It left out ‘norme’ in ‘norme ISO/SAE.’ That’s not optional in French.” 

This wasn’t a complicated technical manual or a legal contract. It was a simple, one-page business document. And yet, the AI translation made mistakes that ranged from cringeworthy to potentially misleading. 

What This Taught Us 

We’re not anti-AI at Rapport International. In fact, I use AI tools every day to streamline tasks and improve productivity — including drafting this very blog. We welcome innovation and are always looking for ways to integrate smart technology into our work. 

But when it comes to translating important information, AI isn’t ready to replace human expertise. 

Because translation isn’t just about matching words — it’s about meaning, nuance, tone, and context. It’s about knowing when “involve” means “participate” and not “implicate.” It’s about understanding that “adding value” needs to sound natural, not robotic or forced. And it’s about making sure critical terminology isn’t lost in translation — literally. 

This experience reinforced a few truths: 

• Not every bilingual person can translate — or even proofread a translation. It takes professional training and deep subject matter expertise. 

• Subtle differences in word choice matter — especially in regulated or high-stakes industries. 

• Cheap translations can cost you — whether it’s in lost revenue, damaged reputation, or legal liability. 

• Consistency matters — working with a dedicated linguist who learns your brand, voice, and preferences over time ensures higher quality and better ROI. 

So yes, keep using AI where it makes sense. But when it comes to communicating clearly and confidently across languages, human translation is still your best bet. 

And please — know who’s doing your translation.