State of the Language Industry 2025: November Update

Man vs. Machine

Last month, after returning from LocWorld54, I decided to take up the pen (which I'm told is mightier than the sword) and write an open letter to the language industry. In it, I addressed the ongoing boxing match between AI optimists and naysayers — a dynamic that has placed our industry in a stranglehold that's becoming increasingly difficult to escape. As another turbulent year comes to an end, I feel a familiar shiver reminding me that I still haven't said everything I wish to say about the matter. As a translator by education speaking four languages, with more than a decade of LSP-side experience and a track record of designing enterprise localization programs, I believe it's my responsibility to keep this conversation going and to address the elephant in the translation room.

Over the last couple of weeks, I spent a considerable amount of time with hundreds of professional translators and interpreters. I spoke on a panel at the European Commission's Translating Europe Forum #2025TEF, facilitated a full-day workshop on AI and MT at KU Leuven, and delivered a presentation and hands-on sessions at the Belgian Chamber of Translators and Interpreters (CBTI-BKVT). I talked openly about AI and emphasized its huge potential as I would in customer-facing scenarios. The feedback was surprisingly positive.

Nonetheless, the year is 2025, and our industry is being held hostage by a race to the bottom. Translation buyers increasingly try to push their entire content stack through machine translation, LSP margins have sunk to historic lows, and professional linguists are paid peanuts to do work they consider an insult to their profession, i.e., post-editing machine-generated content. Makes total sense since linguists are highly educated professionals with the rare ability to communicate across languages and cultures, and combine premium multilingual communication with deep domain knowledge.

The "man vs. machine" debate is fueled by two extremes who view the unstoppable rise of technology exactly the way they want it to be. On one side, corporate stakeholders who, under constant pressure to cut costs at all costs, expect translation to be fully automated, lightning fast, and essentially free. On the other side, human language professionals who cherish crafting meaning through words and rightfully view their work as a form of art. It's a clash between business and passion, ignorance and emotion — a dangerous cocktail for an industry trying to move forward. The consequences are massive:

• It unfairly narrows the conversation around AI down to machine translation.

• While both extremes fail (or refuse) to understand each other's concerns and needs, they add an almost political dimension to the current state of the language industry.

• And most importantly, we lose sight of what truly matters: the enormous gap between AI-only and pure human translation — an entire universe of business models, content streams, and use cases where humans and technology can uplift each other.

According to me, the man vs. machine debate is as artificial as the intelligence that sparked it in the first place. It is completely unnecessary, even irrelevant, and doesn't take us anywhere. Despite all the finger-pointing, it won't produce a winner — not the buy side, not the freelance market, not the think tanks, and also not the kind of monster LSP that's big enough to be the largest TMS provider on this planet, offering many linguists their favorite CAT.

Neural Machine Translation & Translation Memory

Saying that the resilient and crystallized language industry is going through tough times still feels like a euphemism. I don't have magic solutions to our problems, but I firmly believe that the first step toward a better future is understanding the roots of what brought us here.

In my previous article, I gave a name to our industry's pain: neural machine translation. A highly controversial and paradoxical statement, especially since the transformer was introduced by Google specifically for translation before DeepL took it to new heights. The truth of the matter is that we simply weren't ready for a technology as powerful as neural MT — whether in terms of implementation, monetization, or quality measurement.

I'd like to go one step further and drop another provocative opinion. Although our industry thrived after the very successful introduction of the translation memory, it also initiated the gradual undervaluation of human expertise in the translation process. Translation memories, still one of the main commercial assets across LSPs of all sizes, degraded content and translation to units (segmented words), made it possible to quote work upfront without considering complexities such as cultural adaptation or domain-specific terminology, and even standardized discounts for broken translations (indeed, fuzzy matches). The added value that human linguists bring to the table has slowly but surely been sidelined, with per-word rates naturally becoming the primary differentiator to close deals. Like it or not, transferring units from one language to another is something machines can do at a much higher pace and lower cost, which has unavoidably led to the commoditization of translation.

The consequences of this quick history lesson are as obvious as they are hard to deny: while translation buyers were granted the power to push prices down, large LSPs responded with custom development, shiny marketing, and acquisitions. Translators were pulled along by the current.

The Jevons Paradox

The many presentations and workshops I delivered to translators and interpreters the last couple of weeks left me with a very good feeling. It's refreshing to show them that translation technology and AI are about much more than a Trados license, a DeepL API key, and a separate screen for ChatGPT or Copilot. Many linguists thanked me for inspiring them to use AI to make their own life easier. I even heard a senior translator say that despite all the turbulence in the industry, she wouldn't want to go back to the paper dictionary era. Plenty of proof that we're on the right track and, more importantly, that technology isn't exactly the enemy.

I voluntarily choose to mirror my optimism to a very interesting theory, the Jevons Paradox, which entails that as technological improvements increase the efficiency with which a resource is used, the overall consumption of that resource will actually rise — not fall. In the context of AI and translation, the Jevons Paradox suggests that as AI makes translation faster, cheaper, and more accessible, the total demand for translation will paradoxically proliferate and increase rather than decrease overall activity. The future isn't about choosing between man or machine; it's about closing that enormous gap between AI-only and pure human translation, exploring new business opportunities and compensation models, and leveraging AI in all the right places and for all the right purposes to amplify human skill, creativity, and expertise.

Credits to Serhiy Dmytryshyn (Crowdin) for talking to me about these things.