AI Translation Accuracy: What Our 2025 Study Reveals
Discover how accurately today’s leading AI translation tools perform when translating a variety of corporate texts into dominant export languages. We’ve put them to the test in a rigorous, data-driven, in-house study.
Enter your details below to download the white paper.
AI translation has evolved rapidly, but is it reliably accurate for business-critical content?
Artificial Intelligence (AI) is revolutionising the way businesses handle their international communication needs. Translations that once demanded expertise, time and investment can now often be managed in-house, instantly and at minimal cost. But as AI translation tools such as Google Translate, DeepL and Chat GPT become more widespread, a critical question emerges: can AI translation’s accuracy rival that delivered by professional human translators?
AI Translation Accuracy: About the Study
To evaluate how AI translation tools perform in real-world business contexts, AST carried out an 8-week, in-house study focused on measuring AI translation accuracy across dominant export language pairs relevant to professional settings.
For any text that our clients had requested to be translated by AI (using leading AI translation tools), our professional translators were asked to log all errors present in the AI-generated translation. Our translators analysed a wide range of business-critical content types, allowing us to assess how AI performed differently over a variety of texts.
Each error was categorised using a custom error framework – ranging from minor to major and critical – to evaluate not only how frequently errors occurred, but also how significantly they impacted the clarity, resonance and accuracy of the final AI translated text.
To give businesses clearer, more actionable insights, we organised our findings by text type (including legal, technical, corporate & persuasive). This will allow professionals to distinguish where AI translation performs reliably and may be appropriate for their needs, and where it falls short, highlighting when human expertise remains essential.
AI Translation Accuracy: Why it Matters
As businesses increasingly turn to AI tools for fast, low-cost language solutions, understanding AI translation accuracy is more important than ever. While these tools offer impressive speed and accessibility, they can also introduce serious risks when used for complex, business-critical communication.
Inaccurate translations can lead to misunderstandings, compliance issues, brand damage, or lost opportunities – especially in industries where clarity and precision are non-negotiable. Even when meaning has been successfully mirrored, failing to adapt tone, context, or cultural nuances can weaken the impact of your message and reduce engagement with local audiences.
This study gives professionals the insights they need to make informed decisions about using AI translation.
What You'll Learn from the Study
This whitepaper offers a clear, data-driven overview of how today’s AI translation tools perform in real-world business contexts, and where their limitations lie.
You’ll gain insights into:
- The evolution of AI translation tools
- The known shortcomings of AI translation, as per key existing research
- Which content types AI translation performs most accurately on, according to our findings
- Which content types AI translation performs least accurately on, according to our findings
- The key features of each content type that are often lost in ‘AI’ translation, according to our findings
- Which content types are most amenable to human post-editing, according to our findings
- How to make smarter, more strategic decisions about when to use AI translation tools and when to rely on human expertise, according to our data findings
Whether you’re a marketer, an international HR professional, a global marketer, or a product developer, this study will help you understand where AI translation can reliably support your international communication needs, and where it may fall short.
Download the Whitepaper
Make smarter, risk-aware decisions about AI in your multilingual strategy.
Enter your details below to download the whitepaper: a data-driven study developed & executed in-house by expert linguists.
References
- Bhatia, A. & Nicholas, A. ‘Lost in Translation: Large Language Models in Non-English Content Analysis’ (Centre for Democracy & Technology, 2023) Accessed here on 14 May 2025.
- Charles-Kenechi, S. ‘Artificial Intelligence in Translation Studies: Benefits and Challenges’ (Journal of the Department of French & International Studies, 2 (1) (2024)), 5–15 Accessed here on 14 May 2025.
- Hasyim, M. et al. ‘Artificial Intelligence: Machine Translation Accuracy in Translating French-Indonesian Culinary Texts’ (International Journal of Advanced Computer Science and Applications, 12 (3) (2021)) Accessed here on 14 May 2025.
- López-Arroyo, B. & Sanz-Valdivieso, L. ‘Google Translate vs. ChatGPT: Can non-language professionals trust them for specialized translation?’ (paper presented at the International Conference on Human-Informed Translation and Interpreting Technology (HiT-IT 2023), Naples) Accessed here on 14 May 2025.
- Mager, M. et al., ‘Tackling the Low-resource Challenge for Canonical Segmentation’ (2020) Accessed here on 14 May 2025.
- Moneus, A. M. and Y. Sahari, ‘Artificial intelligence and human translation: A contrastive study based on legal texts’, (Heliyon, 10(6) (2024)), e28106 Accessed here on 14 May 2025.
- Wang, J., ‘No Language Left Behind: Scaling Human-Centered Machine Translation’ (2022) Accessed here on 14 May 2025.
- Wang, L. ‘The Impacts and Challenges of Artificial Intelligence Translation Tool on Translation Professionals’ (SHS Web of Conferences, 163 (2023)) Accessed here on 14 May 2025.
- Zhou, F. ‘The Comparison of Translationese in Machine Translation and Human Translation in terms of Translation Relations’ (2022) Accessed here on 14 May 2025.