How Is AI Reshaping the Global Translation Industry – and What Still Needs Humans?

The global translation industry is in the middle of its biggest disruption since the arrival of last generation of CAT tools. Language service providers, publishers and in-house teams are all asking the same question: what does artificial intelligence really change – and how fast?

Market studies show that the wider language services industry was worth roughly USD 72 billion in 2024, after mid-single-digit growth, and is on track to pass USD 90 billion by the end of the decade.

At the same time, AI-driven translation tools have leapt forward. The machine translation market alone reached around USD 1.55 billion in 2023, growing more than 30% in a single year according to Slator’s figures. And with OpenAI’s launch of ChatGPT Translate in January 2026, even more European companies are experimenting with instant, AI-powered translation.

This article looks at how different sub-sectors of translation – literary, corporate, interpreting, media and medical – are being transformed, and why the value of human expertise and the legal framework will shape the future of the profession.

How big is the translation and language services market today?

Depending on the definition, the core translation/localization market is valued at around USD 27 billion (data from market research and consulting firm Nimdzi), while the wider language services ecosystem (including interpreting, media localization and tech) is estimated at about USD 72 billion and expected to pass USD 90 billion by the end of the decade. There are no public and reliable info about sub-sectors market share within the industry.

How are the main sub-sectors of translation affected by AI?

AI is reshaping all translation subsectors—just not in the same way, and not with the same tools.

• Literary translation for books
  • Some European and UK publishers now use ChatGPT, Claude or DeepL for pre-translation, then ask humans to post-edit.
  • Translators report lost work in high-volume genres (romance, cosy mystery, some fantasy/crime, practical guides).
  • Main risks: flattened authorial voice, lower post-editing fees, non-consensual reuse of translations, and a growing split between “prestige” (human) and “midlist/genre” (AI-heavy).
• Corporate translation and localization
  • AI is standard for internal comms, support, product texts and “gisting”, via tools such as DeepL, Google Translate, Microsoft Translator and LLMs.
  • Typical model:
    • Tier 1: low-risk → AI + light check.
    • Tier 2: brand/customer-facing → AI draft + strong human editing.
    • Tier 3: legal, medical, regulatory → human-only translation with CAT/QA tools.
• Interpreting
  • Real-time captions and speech translation are provided by Zoom, Microsoft Teams, Google Meet and other speech-to-speech tools.
  • Good enough for low-stakes internal meetings; insufficient for courts, healthcare, EU institutions or high-level conferences, where human interpreters remain essential.
• Dubbing, subtitling and media localization
  • Whisper, automatic subtitlers, ElevenLabs, Respeecher, Papercup and similar tools automate transcription, first-pass subtitles and synthetic voices.
  • Human experts are still needed for timing, lip-sync, cultural adaptation and rights management, especially around voice cloning and creative choices.

Overall, AI tools are challenging every subsector—but they also make clearer where human creativity, judgement and responsibility cannot be automated.

What does AI change for the translation ecosystem?

Why is a standalone ChatGPT Translate tool significant?

On 15 January 2026, OpenAI quietly launched ChatGPT Translate, a dedicated web-based translation interface linked to ChatGPT.

Key features include:

• a familiar dual-box translator layout, similar to Google Translate;
• support for more than 50 languages out of the box;
• the ability to adjust tone and style (“business-formal”, “friendly”, “for children”, etc.).

This last point is important: unlike traditional MT engines, ChatGPT Translate is designed to:

• interpret context,
• follow instructions on style and audience,
• and generate more natural-sounding target text.

For a marketing manager in Barcelona sending materials to French hospitals, this means they can generate a draft translation, ask for “formal French for healthcare professionals”, and get something that feels reasonably polished – at least on the surface.

Where does AI translation still fall short of human quality?

Despite striking progress, expert reviews and industry tests still highlight several weaknesses:

• complex, culture-dependent texts (literary, humorous, strongly idiomatic);
• highly specialised domains such as regulatory, legal or medical translation, where:
  • terminology is dense and constantly evolving,
  • errors can have legal or clinical consequences;
• under-resourced language pairs, or domain-specific jargon where the model has seen little high-quality data.

According to Slator, the machine translation market has grown rapidly, reaching USD 1.55 billion in 2023, a 31% increase compared to 2022. While it still represents a relatively small share of the overall translation market, it is expected to continue growing at a very fast pace. AI is already powerful enough to significantly transform workflows, but it is not yet reliable enough to fully replace human responsibility, especially in contexts where people’s health, financial interests, or legal rights are at stake.

What new opportunities can AI create for translators and clients?

Used intelligently, AI can expand the role of translators instead of shrinking it. For example:

• Productivity gains for repetitive or low-risk content.
• Ability to test messages quickly across multiple languages before investing in full localisation.
• New services like:
  • multilingual content audits,
  • terminology mining using AI,
  • consulting on AI/MT strategy for global companies.

For linguists, this shifts the centre of gravity from “I type every word myself” to “I design and supervise a high-quality multilingual workflow”:

• designing prompts and guardrails,
• auditing AI output,
• providing high-value human adaptation where it matters.

How will the value placed on human work reshape the translation industry?

Why does “handcrafted” translation still matter in a machine age?

In a machine age where people across the world happily use AI to save time and money, there’s still a deep, almost instinctive respect for “artistic” work born of human skill—whether in literature, photography, painting or other crafts—because we recognise the sensitivity, know-how and lived experience behind it, and translation increasingly belongs in that same category of human hand-crafted work.

In this context, many readers and clients intuitively feel that:

• a human translator brings a unique mix of knowledge, ethics and creativity;
• a text produced by a machine, no matter how fluent, is not equivalent to human expression.

This is particularly visible in:

• literature and non-fiction (authors often insist on human translation);
• brand-critical content (slogans, campaigns, investor communication);
• sensitive domains such as medical or legal, where responsibility must be clearly human.

The challenge for LSPs and freelancers is therefore to make this added value visible both in marketing and in pricing.

How can we clearly differentiate human and AI-assisted translation?

One practical step is radical transparency:

• Clearly label services as, for example:
  • “Human translation & independent revision”
  • “AI-assisted translation + expert editing”
  • “Raw AI output (not recommended for external use)”
• For books and cultural products, industry bodies and translator collectives like Against Writoids are calling for:
  • clear disclosure when AI is used at any stage;
  • no public funding for AI-generated books or translations presented as original creative work;
  • maintaining copyright and moral rights for human authors and translators.

This kind of labelling is good not only for ethics, but also for SEO and GEO marketing: Companies can highlight “human-reviewed medical translation” or “AI-assisted, human-certified technical translation” as distinct service lines.

How could legal and regulatory frameworks shape the future of AI translation?

As AI translation becomes standard in many workflows, laws and regulations will increasingly decide what is allowed, who is protected and who is responsible when things go wrong. Legal and cultural organisations are already signalling where the pressure points will be.

Why are legal and cultural organisations worried about AI in translation?

Many authors’, translators’ and cultural organisations warn that unchecked AI use in translation could:

• blur or undermine intellectual property rights for authors and translators;
• weaken contracts and moral rights, especially when human work is treated as cheap training data;
• have social and psychological impacts, if creative jobs are systematically replaced by machine-generated content;
• increase environmental costs, given the energy required to train and run large AI models.

To counter this, they are calling for:

• strong transparency on AI-generated books and translations (clear labelling, clear workflows);
• rules for public cultural funding, so it supports human creativity rather than mass AI-generated output;
• robust compensation mechanisms when human translations or texts are used to train AI systems.

If these demands are reflected in law, they could slow purely cost-driven AI deployment and push the market towards more ethical, human-centred models.

How could AI laws and regulations affect translation and language technologies?

New AI laws – such as the EU AI Act and similar frameworks being discussed worldwide – tend to follow a risk-based approach that will directly affect translation use cases:

• Minimal-risk AI (most everyday tools) may face lighter rules but still need basic transparency.
• High-risk systems used in medical, legal, safety or employment contexts could face strict obligations on:
  • data quality and documentation,
  • transparency about model capabilities and limits,
  • human oversight, robustness and incident reporting.
• Providers of general-purpose or foundation models used for translation may be required to:
  • document training data and respect copyright,
  • disclose when and how AI is involved,
  • implement risk-management processes.

For organisations using AI translation in healthcare, legal, financial or other regulated sectors, this could mean:

• clearer documentation of AI pipelines and decision points;
• mandatory human review and sign-off for critical content;
• stricter data protection rules, especially around personal and sensitive information.

In practice, regulation is likely to make “AI-only” translation harder to justify for high-risk content, while encouraging hybrid, auditable workflows that combine AI speed with human responsibility.

Why is responsibility likely to remain with humans, not machines?

One core legal question is: who is accountable when an AI translation is wrong or harmful?

Because current and emerging laws treat AI as a tool, not a legal agent, responsibility will continue to rest with:

• the company that chooses and deploys the AI system;
• the language service provider (LSP) that builds AI into its workflows;
• the human professionals who validate and approve the final text.

If a mistranslation affects a contract, a medical document or a safety instruction, regulators will still look for a human or organisation to hold accountable.

This is a key reason why, in high-risk areas, fully automated AI translation is unlikely to become the norm: legal and regulatory frameworks will keep pushing businesses back towards traceable, human-supervised translation processes.

How does AbroadLink approach AI in medical translation?

How has AbroadLink been using translation technology long before today’s AI hype?

At AbroadLink, technology is not new: the company has been working with CAT tools, translation memories and terminology databases for many years;

The philosophy is simple:

Use technology where it adds value, never where it might compromise safety or compliance.

Why is human intervention non-negotiable in medical and life sciences translation?

Life sciences and medical translation, including the translation of medical devices, are different because:

• documents are densely technical and heavily regulated;
• terminology must align with official sources (EMA, EDQM, MedDRA, national agencies);
• errors can directly affect patient safety or regulatory approval.

That’s why AbroadLink’s workflows for:

• IFUs and manuals,
• labelling and packaging,
• clinical trial documentation,
• pharmaceutical brochures

always include human specialist translators and independent reviewers, even when AI has been used earlier in the chain to accelerate specific tasks.

How can clients build an AI-aware but risk-aware translation strategy with AbroadLink?

For medical and life sciences companies, AbroadLink can help:

• audit multilingual content and classify it by risk;
• design a tiered translation policy (human-only vs AI-assisted);
• integrate custom generative translation where appropriate, always with medical post-editing;
• document processes in a way that supports regulatory compliance and future audits.

This lets clients benefit from faster turnaround and better cost control, without losing sight of the essential: accuracy, traceability and safety of translation for the medtech industry .

Conclusion

AI is expanding at high speed, and like many other sectors, translation is feeling the impact in every corner of its ecosystem. Yet the attachment to human know-how, creative skills and clear human responsibility—combined with emerging legal safeguards—will help preserve crucial areas where machines cannot fully replace people. For organisations, the safest and most efficient path is to rely on translation companies who can combine the best of AI with expert human translation, ensuring both innovation and trust.