AI Workflows Outperform Human Translators in China Study

Image credit: Search Engine Journal

AI-driven localization workflows outperformed human translators in most content categories for English-to-Chinese translation, according to a new benchmark study released in China.

The research indicated that artificial intelligence workflows scored higher than human translators in four of six content types, challenging traditional professional translation services.

Human translators did not rank among the top five performers for technical, product user interface (UI), user-generated content (UGC), and marketing materials. For marketing copy, human translators finished 10th, trailing the leading AI workflow by 22.2 points, the study reported.

Conversely, human translators achieved higher scores in informational and search engine optimization (SEO) content. However, their lead over the best machine workflow was narrow, only 2.8 points for SEO content.

The study, a joint initiative by EC Innovations and Jademond Digital, evaluated 774 localized outputs. These outputs spanned six distinct content types, seven different workflow models, and three task types.

Chinese native-speaking professional localizers blindly scored all outputs, ensuring an objective assessment, according to the organizations.

Both EC Innovations and Jademond Digital are companies that provide related localization and digital services.

The findings suggest a significant shift in professional translation, particularly for high-volume and specialized content types, where AI is demonstrating increased proficiency.

Researchers said the comprehensive nature of the study, covering a wide range of content and workflow variations, provided data on the current capabilities of AI in localization.

The results underscore the ongoing evolution of AI technologies and their increasing impact on industries traditionally reliant on human expertise.

Source: Search Engine Journal