Advancements in artificial intelligence (AI) have significantly altered the landscape of professional translation, creating new challenges for human translators. Whereas some fields, such as software development, have seen AI improve the quality and satisfaction of work, translators report that the incorporation of AI has fragmented and mechanized their jobs, often to their detriment.

Traditionally, translators were engaged in specialist, knowledge-intensive work involving the interpretation and creative rendering of meaning from one language to another. However, with the rise of machine translation tools and AI-powered systems, many translation agencies have shifted to a model known as machine translation post-editing (MTPE). Under this system, machines produce initial translations, which human freelancers are then tasked to review and correct. This shift has led agencies to reduce pay rates sharply, sometimes by as much as two-thirds. For example, one translator, Petr Čermoch, who works on TV subtitles from English to Czech, noted that a major agency lowered its rate from $5 to $1.50 per minute of video following the adoption of MTPE. Similar pay reductions have been reported in legal, financial, and academic translation sectors involving other language pairs.

Translators describe MTPE as cognitively demanding and more complex than working from scratch, as it requires closely comparing the machine-generated text with the original source language to ensure accuracy and quality. Mark Rawson, who translates between English and Chinese, said thoroughly checking machine output involves intense scrutiny of both versions of the text, making it two to three times more difficult than pure translation work. Despite this increased effort, rates have been reduced substantially, forcing translators to work faster to maintain their income. Many report that this accelerated pace, combined with the need for meticulous quality control, generates significant stress and fatigue.

The emerging work model is also seen as less fulfilling. Čermoch described the job under the MTPE system as mechanical and bereft of the creative aspects he once enjoyed. The character of translation has shifted from a nuanced craft to a routine task, diminishing its appeal to experienced professionals. Consequently, some seasoned translators are opting out of MTPE work altogether. As they exit, less experienced translators are taking on these roles, and some companies are beginning to rely entirely on machine translations without human post-editing.

Labor market data from the United States corroborates these individual accounts. The rapid growth in translator and interpreter employment slowed significantly around 2010, coinciding with the rise of Google Translate. Since then, the relative wages of translators have declined, and the number of translators employed has fallen by nearly 20 percent from its peak in recent years. New translation projects advertised on online freelance platforms have dropped by nearly half in the two years following the launch of AI language models like ChatGPT. While such data primarily reflects employment rather than freelance activity, the freelance segment likely faces even more pronounced effects due to less job security and regulation.

The experience of translators contrasts strongly with that of software developers, where AI tools have automated routine coding tasks, allowing developers to focus more on collaborative and creative aspects of their jobs. In translation, by contrast, AI has superseded the creative core of the work, relegating human involvement to slower, more error-prone post-editing of machine output.

This divergence highlights how the impact of AI depends critically on which elements of a job are automated and how that automation influences the role of human workers. Translation’s transformation raises broader questions about whether its trajectory represents a unique case, given its close alignment with machine language models, or if similar patterns may soon affect other knowledge-based professions.