When I started working in the translation and localization industry more than two decades ago, the idea of machine-generated language felt like science fiction. Automated translation existed, but it wasn’t any good. Back then, our work was unquestionably human. We were, as we continue to be, gatekeepers of tone and nuance. Where machine translation (MT) was available at all, it produced rigid, awkwardly mirrored sentences with no grasp of intent or tone. As a linguist, it wasn’t something you feared, it was merely something you fixed.
Fast forward to today, and AI has completely transformed the translation landscape. Tools that once butchered meaning are now producing eerily fluent copy. Large language models (LLMs) are capable of translating long, complex chunks of text with extraordinary accuracy. Computer-assisted translation (CAT) tools now come with integrated neural engines. The pace of change is overwhelming, and understandably, many colleagues and players in the industry are wondering where this leaves us humans.
Believe me, I understand the concern. But from personal experience, I’ve also learned this: AI hasn’t made me obsolete — it’s made me more valuable, especially when I embrace it on my own terms.
I’ve lived through every stage of this evolution. I started working in the subtitling industry circa 2000, then became a Sworn Translator, and ended up founding a boutique translation agency. Over the years, I’ve seen firsthand how large organizations have shifted from purely human localization pipelines to hybrid models, driven by the growing promise of AI.
But let’s be clear: It hasn’t always worked smoothly.
Years ago, I was tasked with overseeing a large-scale localization project for a major tech giant. The project involved training a voice-enabled assistant using user-generated content pulled from high-traffic online forums. The source material was already difficult, fragmented, slang-heavy, and culturally specific. The MT output? Even worse. It delivered flat, literal renderings of content that made no sense out of context.
My team and I spent hours interpreting the intent behind these posts, researching cultural equivalents, and rewriting everything from scratch to make it relevant for Latin American audiences. In that scenario, MT wasn’t just unhelpful, it was a barrier we had to work around. And this was from a proprietary internal MT engine touted as being more powerful than the public tool that was available at the time.
What’s changed since then? A lot!!!
Around 2018, I started noticing a shift. As voice-enabled virtual assistants went mainstream, so did the research driving their language models. Translation tools began to pick up more context. Segmentation became smarter. Syntax improved. When LLMs like ChatGPT emerged, the pace of evolution skyrocketed. We moved from MT systems that handled sentences in fragments to models that could generate cohesive, human-sounding paragraphs.
At the same time, CAT tools evolved. Platforms like Phrase and Smartling, once limited to leveraging termbases and translation memories, began to integrate AI-powered MT. And when configured properly (which is absolutely essential), these tools started producing drafts that were genuinely useful. Today, with the right segmentation, prompts, and review workflows, AI can become a powerful asset to human localization teams. Not a replacement, but an amplifier.
That said, the risks haven’t vanished. AI still stumbles over idiomatic expressions, humor, emotion, and regional nuance. It still misses cultural cues that only a human translator can catch. It still struggles to replicate tone, irony, and subtext. And it certainly doesn’t understand a brand’s voice or mission unless someone properly trains it to — and that’s not an easy feat.
What all this means is that the role of the translator has evolved and continues to evolve, but it hasn’t and won’t disappear. We are becoming editors, strategists, cultural consultants, and, increasingly, the human voice guiding AI toward better output. The craft is still there, it’s just shifting upstream.
Translation is no longer just about words, but about intent, emotion, positioning, and connecting. AI can generate content, but it can’t yet interpret values. It can mimic language, but it can’t speak to people’s realities. That’s where we human translators, editors, and quality assurance specialists come in.
I’ve learned that translators who adapt and learn how to prompt, curate, post-edit, and quality-assure are not being pushed out of the profession. Instead, we’re rising to the top of it. That’s why I believe AI hasn’t replaced me — it’s made me better at what I do!