AI and Translators: What's Actually Happening and What to Do

The honest assessment

Translation has experienced one of the most dramatic AI transformations of any profession. The quality leap in machine translation over the past few years hasn't been incremental. It's been seismic. And if you're a professional translator, you already know this because you've watched it happen in real time.

Google Translate made a quantum jump in 2016 when it switched to neural machine translation. DeepL launched and quickly became the preferred tool for many professional translators because its output was, frankly, better than many human translators for straightforward text. Then generative AI arrived. ChatGPT, Claude, and Gemini don't just translate word by word or even sentence by sentence. They understand context, tone, register, and intent. They can translate a marketing brochure into something that actually reads like a marketing brochure in the target language, not a translated document. For many common language pairs and text types, AI translation is now genuinely good.

The market impact has been severe. The European Commission's 2024 study on the language industry found that demand for basic translation services has declined significantly. Corporate translation budgets are being redirected from human translators to AI translation with human post-editing. Some translation agencies have restructured their entire business model around AI translation with human quality assurance rather than human translation with technology assistance. Rates for standard translation work have dropped across most language pairs.

But here's where the picture gets more nuanced. Not all translation is created equal. Translating a user manual from German to English is a very different task from translating a novel. Translating a legal contract is different from translating a marketing campaign. Translating a medical journal article is different from translating a tourism website. AI handles the functional end of this spectrum well. It handles the creative, specialist, and high-stakes end less well. The closer the work is to "conveying meaning between languages," the more automatable it is. The closer it is to "recreating experience, intent, and cultural resonance across languages," the more human it remains.

What AI still gets wrong: cultural nuance that requires lived experience in both cultures. Humour that relies on wordplay, cultural references, or double meanings. Legal precision where a single word choice could change the legal interpretation. Literary style where the author's voice needs to be preserved, not just their meaning. Marketing copy that needs to persuade in the target culture, not just communicate. These are the areas where human translators remain essential, and they're also the areas that command the highest rates.

Your exposure level: High

High. The broad translation market is being fundamentally restructured by AI.

The core problem is that translation, at its most basic, is a pattern-matching task between two languages. Given sufficient training data — and for major language pairs there are billions of translated sentence pairs available — AI can learn these patterns extremely well. The European Association for Machine Translation has documented steady improvements in machine translation quality scores, with neural machine translation closing the gap with human translation for many text types.

The economic pressure is immense. A professional human translator in the UK charges roughly £0.10-£0.15 per word. AI translation costs a fraction of a penny per word. Even with human post-editing adding cost, the total is typically 40-60% less than fully human translation. For organisations that translate millions of words per year — large corporations, EU institutions, software companies — the savings are enormous. The shift from "translate from scratch" to "post-edit machine translation" is not a trend. It's the new industry standard.

The exposure is somewhat lower for rare language pairs (where AI has less training data), highly specialised domains (where accuracy is critical and domain expertise matters), and creative translation work (where cultural adaptation is the primary value). But for the bread-and-butter work that sustained many translators' incomes — general business translation, technical documentation, website localisation — the market has shifted permanently.

The 90-day action plan

1. This week: honestly evaluate your translation work against AI output. Take a recent translation project. Run the source text through ChatGPT, Claude, and DeepL. Compare the outputs to your work. Where is your translation genuinely better? Where is it comparable? Be ruthlessly honest. The gaps you identify are either areas where you add clear value (focus on these) or areas where you're doing work AI can handle (let it).

2. Week two: master the post-editing workflow. MTPE (Machine Translation Post-Editing) is the dominant model now. If you're not already doing it, learn to edit AI translations efficiently. This requires different skills from translating from scratch — it's about spotting and correcting errors rather than producing from blank. It's less creatively satisfying, but it's the market reality. Getting fast and accurate at post-editing keeps you employable while you develop higher-value skills.

3. By day 30: develop a deep specialism. Legal translation. Medical translation. Literary translation. Financial translation. Patent translation. The more specialised and high-stakes the domain, the more human expertise matters. If you're a generalist translator, you're competing directly with AI. If you're a specialist translator with domain expertise, you're providing something AI can support but not replace. Pick a domain and go deep.

4. By day 45: build your transcreation and cultural adaptation skills. Transcreation — creative translation that adapts content for a target culture rather than just a target language — is the area of translation most resistant to AI. Marketing campaigns, brand messaging, entertainment content, and cultural communication all need transcreation. This is where translators become cultural consultants, and it commands premium rates.

5. By day 60: learn the technology ecosystem. Understand CAT tools (SDL Trados, memoQ, Memsource), translation management systems, and how AI is integrated into professional translation workflows. Being fluent in translation technology makes you more efficient and more attractive to agencies and direct clients. The translator who can manage technology is more valuable than the one who just translates.

6. By day 75: develop adjacent skills that compound with translation. SEO in multiple languages. International marketing strategy. Cross-cultural communication consulting. Localization project management. The translator who brings strategic value beyond the translation itself is operating at a level AI can't reach. You understand cultures in a way that a language model doesn't.

7. By day 90: reposition your professional offering. Stop selling "translation" and start selling "cross-cultural communication." Update your profiles, your website, and your pitches to reflect the strategic value you provide. "I help organisations communicate effectively across cultures" is a very different proposition from "I translate English to French." The first is hard to automate. The second isn't.

AI tools you should be using this week

• ChatGPT for Work — Use it as a first-draft translation engine for appropriate text types, then apply your expertise to refine the output. Also invaluable for terminology research, checking idiomatic expressions in context, and exploring alternative phrasings. When a client asks for a translation in a sub-dialect or register you're less familiar with, ChatGPT can help you identify the right tone.

• Claude for Work — Particularly strong for long-form and complex translations where maintaining consistency across the document matters. Claude handles context well, which means it can maintain terminology consistency and tone across a lengthy text. Use it for first drafts of reports, white papers, and legal documents, then apply your domain expertise to correct and refine.

• Google Gemini for Work — Google's AI has access to vast multilingual data and handles many language pairs well, including some less common ones. Useful as a comparison tool — run text through multiple AI translators and compare outputs to identify the best starting point. Also strong for quick terminology checks and understanding usage patterns in the target language.

• Grammarly AI — Use it on your target-language output (for English targets) to check natural flow, tone, and idiom. AI-translated text that's been post-edited can sometimes retain subtle non-native patterns that Grammarly catches. It's a final quality check that helps ensure your output reads naturally.

What to say in meetings

When clients ask if they should just use AI translation: "For your internal documentation and basic communications, AI translation with a quick review is probably fine and I'd recommend it. For anything customer-facing, legally binding, or brand-sensitive, you need a human translator who understands your industry and your target audience. I use AI as part of my workflow, which means you get speed plus quality assurance. What you don't get from AI alone is the cultural understanding and domain expertise that prevents embarrassing or costly mistakes."

If agencies are pushing post-editing at reduced rates: "I understand the market is shifting. I'm happy to include post-editing in my workflow where appropriate. However, my value isn't just in the words — it's in the domain expertise and cultural understanding I bring. For [specialist area], the cost of a mistranslation far exceeds the savings from AI. I'm pricing for the expertise, not just the word count."

In professional development conversations: "I've been developing my specialism in [domain] and my transcreation capabilities. I've also integrated AI tools into my workflow so I can deliver faster without compromising quality. My focus is on the work that requires genuine cultural and domain expertise — the work AI can support but can't do alone."

If the worst happens

If translation work dries up in your language pair or specialism, your skills open doors to more roles than you might expect. You're bilingual or multilingual. You understand multiple cultures deeply. You can communicate complex ideas clearly. You have attention to detail that's been tested on every word of every document you've ever translated. These skills transfer to international business development, global marketing, cross-cultural consulting, diplomatic roles, language teaching, and any organisation that operates across borders.

Adjacent roles to consider: localisation manager, international marketing specialist, cultural consultant, language technology specialist (working with the AI tools rather than competing against them), interpreter (which requires real-time human presence), subtitler for specialist content, language teacher, or international communications manager. Many translators also build successful careers in quality assurance for AI translation systems — who better to evaluate and improve machine translation than someone who deeply understands both languages?

Here's what i want you to take away from this. The era of translation as primarily a linguistic exercise is ending. The era of translation as cultural expertise, domain specialisation, and communication strategy is beginning. AI handles the linguistics increasingly well. What it can't handle is the deep understanding of what a message means in one culture and how to make it resonate in another. That's your expertise. Build on it, specialise it, and price it accordingly. The translators who do this will earn more, not less, in the AI era. The ones who try to compete with AI on word-for-word translation will not.