CIOL Council Member Emma Gledhill reflects on thirty years in the languages profession which, as she says, “teaches you a few things.”

One is that whenever translation is reported to be on life support, it's actually just changing. And you can change with it.

 

Jobs are Changing, Not Disappearing

---

Over the last thirty years, the profession has seen three major technological revolutions:

• Translation memory
• Neural machine translation
• Generative AI and large language models

The latter dominates the conversation today and is the most disruptive of the three by far. CIOL Council has tracked the progression of AI at a systemic level. We conclude AI is not going away, and it can't be regulated away, so it’s important to determine what judicious use looks like.

The questions now are often practical ones: how to respond to the revolution, where and when to use the new tools available to us (and when not to) and how to use them safely and responsibly when you do.

 

There is a Future – Just Not the Same One

---

Straightforward, general and highly standardised work in well-resourced languages is increasingly a territory geared to AI and machine translation. 

As Zoe Green MCIL, a highly experienced IT translator and technical writer based in Rome said to me:

"The days have gone where I used to get really easy translations that anyone could do standing on their head." 

What she gets now is highly specialist, terminology-heavy and complex. That's a market signal, and it points directly to why human expertise is still crucial.

Two small but everyday examples from the same week in mid-2026 provide a perfect practical illustration.

 

Example One: Humans Effortlessly “Repair”, Machines Often Can’t 

---

The first came from a vendor on LinkedIn evangelising about their AI-powered LQA tool, which had flagged an inconsistent switch between formal and informal "you" forms across paragraphs of a translation. That translation can only have been generated by a machine because a human would never have mixed up "you" forms in a text in the first place.

The AI or MT created an inconsistency that a human would have repaired in their head before it ever hit the page. In other words, machine-generated translation manufactured a needless repair job and then required another machine to clean up after it. That's not to say a second process is not necessary – four-eyes checks have always been standard and are recommended in ISO 17100. But what value getting it right first time?

 

Example Two: Context is Everything and Machines Don’t Have It

---

The second example was in our own work: a small localization job of five short strings, fewer than ten words between them, covering a functionality that we as users see in almost every app on our phones every day. Machine pre-translation generated major errors in two or three of the five strings, ranging from the merely amusing to the offensive. 

What made the errors easier to catch was context: for once, the strings sat close together, and the key names offered more clues. This information combined with our extensive professional experience – and just life experience – allowed our brains to compile a picture of what all those disparate segments were doing, what we as users would expect to see and how we would expect the functionality to work. Combining translation, localization and UX skills in one thought process. Yet the machine translation got the other half of the strings right and saved time.

Two things can be true at the same time.

 

Adoption Outstrips Understanding

---

These very recent examples illustrate exactly what analysis from industry publication Slator pointed to as a feature in 2025: AI adoption in the language industry is still outstripping understanding and maturity of the technology. 

Trust, quality and risk concerns remain. Judging the boundary between where it helps and where it creates more work is itself a skill, one that only an experienced professional linguist can bring to the table. 

It's translation skills that allow linguistic repair to happen on the fly. It's broader knowledge and lateral thinking that allows us to walk in the user's shoes – because every one of us IS a user. And it's a working knowledge of AI's strengths and weaknesses that enables you as a professional to focus on likely problem areas.

That skillset is your foundation. The question is what you build on top of it.

 

Extend the T, but Keep the Stem

---

CIOL CEO John Worne talked to me recently about the concept of "T-shaped people". T-shaped people have a deep “stem” of specialist competence but add a broadening transversal skillset across the top. 

It's a useful model but daunting to do sometimes: when experienced professionals step outside their core expertise, “imposter syndrome” can raise its head. The fear is often that professionals in other fields might ask you what qualifies you to have an opinion. And going too broad for too long risks you losing confidence in your deepest skill and historic “differentiator” entirely.

There is another aspect John flagged to me, that often goes unspoken: personal identity. Deep specialists (and translators are among the deepest) can risk losing their sense of professional self-image when they expand their offering.

 

Double Down on Skills?

---

For me the answer has been upskilling, and not just free online courses (although those are certainly valuable too if you don’t have the time or money to take a formal qualification). I think paying for rigorous, formal training from a reputable institution sends important signals: that you're serious about investing in yourself, that the qualification carries weight, and that your broader skillset has been independently verified, not just self-declared. 

Experienced translators who have focused their continuous professional development purely within translation, however excellent their skills, may find it harder to stay competitive. 

Translation graduates who moved directly into project management without building hands-on translation experience are capable organisers but often lack practice in the craft. People who genuinely stand out in today’s languages market have often grown in multiple directions – technical skills, UX, technical writing, etc. – and can demonstrate this growth with qualifications.

The rare person who has deep translation experience and a broader, verified skillset is extraordinarily hard to find. And that's an opportunity. 

 

The Market is Moving – Are We?

---

As the scope of language roles expands into localization, content marketing, cultural adaptation, UX writing, consulting, AI supervision and beyond, translation skills become the foundation, not the finished structure. The “portfolio career” and diversification of income stream that many linguists are building, consciously or not, is the natural extension of that foundation.

Diversification of skills amid a convergence of domains may sound like a paradox. However, it’s not about abandoning what we know; it’s about deploying that knowledge across as many areas as it can credibly support.

 

Zoe Green’s Path

---

Zoe Green's path illustrates how this can happen organically. She applied to an IT translation agency and was told they had no need for translators – but would she be interested in technical writing? That contract has now run for over 18 months. At 49, working full-time and with a seven-year-old at home, she has gone back to university to study computer science. 

She's finding that translators have an instinct for code that others don't: programming languages are called languages for a reason. Understanding syntax, capitalisation rules and the consequences of misplaced punctuation turns out to transfer directly.

 

My own trajectory

---

Hired as a translator, reviewer and terminologist over six years ago, my role has since diversified to almost 100% localization, strategy and governance. 

I have gone back to study part-time on the TCLoc master's degree at the University of Strasbourg. Its curriculum has given me the technical understanding to work directly with designers, business analysts and developers; to bridge the gap between linguists and engineers; to understand some of the psychology of how users interact with software and other content (of which language is a symbiotic part); and to identify not just that a localizability problem exists, but why it's a problem and what needs addressing. 

 

Redefining Ourselves?

---

Chris Raulf, international AI SEO and GEO expert and founder of Boulder SEO Marketing, who started his career with a decade in localization, observes that the people struggling most are the ones who defined themselves by the task, while the ones thriving have redefined themselves by the expertise underneath that task: the same underlying skill, deployed across more ground. Translation isn't disappearing, it's transforming into judgement work.

Many working linguists today find it difficult to advise a school-leaver or new graduate to train in translation in the current doom-laden, AI-dominated climate. Yet thirty years of hands-on experience keeps proving its worth to people already in the profession. 

As Zoe Green put it to me: 

“You're never too old to adapt, and you've got a better chance of succeeding if you try than if you don't try." 

No amount of AI progress changes what that experience is built from: doing the work hundreds and hundreds of times until judgement becomes instinct.  Although AI can replicate output, it cannot replicate the judgement that tells you whether the output is right or wrong. 

And that's why the news of translation's demise is decidedly premature!

 

 

Emma Gledhill FCIL CL is Head of Localization at Avaloq and an experienced business linguist working from German, Dutch and French into English, specialising in localization, UX, governance and strategy. A long standing CIOL member, Council member and Fellow, she is active as a conference speaker, panellist and is Chair of the CIOL Membership Committee.

---

This post is an expanded version of the original published on the University of Strasbourg TCLoc blog here 

---

Views expressed on CIOL Voices are those of the writer and may not represent those of the wider membership or CIOL.