Artificial intelligence has transformed the localisation industry at an extraordinary pace, giving organisations access to powerful tools that produce multilingual content in seconds, at lower cost and with faster turnaround than previously possible. However, alongside the excitement sits differing levels of misinformation and misunderstanding. Can AI replace human translation? Is it secure? How do we ensure quality? The reality is more nuanced than the headlines suggest: the most successful organisations aren’t replacing people with technology – they’re combining the strengths of both.
AI is changing far more than translation
AI localisation now spans machine translation, AI voices, transcription, subtitling and real-time interpreting. Some applications are already strong – AI-generated voiceovers sound remarkably natural, and modern AI localisation models produce high-quality first-draft translations for many language pairs. Others remain less mature: transcription depends heavily on audio quality and clarity, and live AI interpreting, while improving, can still compound errors across transcription, translation and speech generation. Understanding where AI performs well matters just as much as understanding its limits.
Speed and savings are real – but so are the risks
For suitable content, AI-assisted localisation combined with professional human post-editing (MTPE) can reduce costs and turnaround by around 30%-40% compared with human translation alone – particularly valuable for high-volume, frequently updated or multilingual content. But efficiency shouldn’t come at the expense of quality or security. Free or consumer-grade tools may retain uploaded content, use it for model training or transfer data across jurisdictions – a genuine concern for regulated sectors and content containing sensitive data. Before adopting AI, organisations should classify their content, understand data sensitivity and use enterprise-grade platforms with proper security controls and data residency options for sensitive or confidential content. It is key to check what happens in the background with your data once it is uploaded into a platform – how it could be used and where it will be transmitted. A professional language partner will have carried out these checks and audits before onboarding the tools they use, so you can be sure that any data they process using these tools is compliant with data security guidelines, but it’s always best to check with them what data security processes they work with.
Why AI still needs human expertise
Modern LLMs have significantly improved translation quality, but they’re not perfect. In order to achieve content that is a comparable level to human quality, all copy (particularly creative, culturally sensitive, legal and highly technical content) still needs some form of professional linguistic review. This is also particularly true of content from regulated industries, where quality assurance checks are paramount. LLMs can also “hallucinate” – adding, omitting, or subtly changing meaning in ways that aren’t always obvious. Human post-editing remains essential because linguists don’t just correct grammar; they validate terminology, protect brand voice, ensure compliance and confirm the translation says exactly what the source intended. The right question isn’t whether AI can replace translators, but how human expertise can maximise the value of AI.
Quality depends on more than the engine
Translation Memories, approved terminology, style guides and good file preparation all shape output quality as much as the AI engine itself. New technologies now allow output to be tailored to company-specific style and terminology through glossaries and reference material. This, in turn, allows for higher levels of quality from the automated output, resulting in quicker post-editing times and a final output that is more aligned with expectations.
Integrated AI tools
Many content management tools include an integrated AI Localisation option, allowing content located within the respective system to be translated in-situ almost instantly. This form of integration is extremely useful where no post-editing or ‘training’ of the automated translations is required, as these tools do not often allow for this functionality. However, where post-editing is required and/or the AI engine needs ‘training’ with corporate language, it may be worth considering a manual integration to a tool that allows these options. In short, integrating AI translation directly into content management systems can boost speed, but without proper terminology management, QA and post-editing, time saved through automation is often lost again to poor quality, corrections and inconsistencies.
The evolving role of the LSP
As AI becomes more accessible, the role of a language services provider is evolving, not disappearing. A good Language Service Provider will have a suite of AI technologies from which they can select to suit different content types and languages. Many AI engines produce better results with certain types of content and languages than others. As such, the role of many LSPs is evolving to focus on selecting the best technology solution and workflow, file preparation for optimal machine translation, terminology and Translation Memory management and coordination of expert review – all while remaining technology agnostic, choosing the best solution per project rather than being tied to one platform.
Looking ahead
AI will keep improving, and developments such as automated quality estimation and automated post-editing will reduce manual intervention even further. Human expertise, however, will remain central to high-quality localisation by focussing on measures to improve AI output and refining content even further for optimum quality. The greatest returns will go to organisations building intelligent hybrid workflows – not pursuing full automation. The future of localisation isn’t AI versus humans. It’s AI, guided by human expertise.
