1 Trillion Words a Month: Why Translators Still Matter in AI Era

Machine translation now handles roughly one trillion words a month through Google Translate alone, according to figures Google disclosed in April and reported by BBC News. That volume does not include DeepL, Microsoft, Claude, ChatGPT, Gemini, or custom corporate translation systems, all of which now compete for the same work that once went to human linguists.

The shift is forcing a narrower question: not whether AI can translate, but where a human still has to check it.

How does idiom trip up AI translation?

François Yvon, a research director specializing in automatic translation at the French National Centre for Scientific Research, points to the French phrase "il pleut des cordes" — literally "it's raining ropes," meaning a downpour. A well-trained system will usually render that into English as "it's raining cats and dogs," Yvon told the BBC. "But is it going to find the equivalent when it's translating into Indonesian?" he asked. A weather idiom is low-risk. Legal filings and medical records are not.

Where does AI translation go wrong on legal documents?

Ariella Baltié, a translator and legal expert who translates court documents from English into French, described a recent case in which a client had already run his paperwork through AI and asked her to check it. "There were two major errors right at the beginning," she said. The system had translated "le demandeur" — the plaintiff — as "the requester," an error she called "totally awkward for legal papers."

Why does Wikipedia still pay humans to translate?

John Zimmerman, founder of the Swiss non-profit OKA (Open Knowledge Association), runs what he says is the single largest contributor of translated articles to Wikipedia, with more than 10,000 pieces published, mostly into Spanish, Portuguese and English. The group could not have reached that scale without AI. "Translating 10,000 articles the traditional way would cost millions," Zimmerman said. "In our process, the AI produces the draft, then a human translator checks every sentence."

He argues the calculation differs by stakes. "For plenty of use cases, raw AI translation is now the right call: multiplying your costs by 100 to catch the last few percent of errors isn't rational if you're translating internal documents or product listings," Zimmerman said. "For an encyclopedia it's different. Wikipedia runs on trust. A wrong fact can sit there for years and get cited by journalists and AI systems."

Is skipping the human translator now realistic?

Zimmerman says the technology has closed the gap fast. "Two or three years ago, with tools like DeepL, our translators still rewrote a good chunk of every draft. Today the prose often doesn't need fixing at all," he said. "So the question is no longer whether AI translation is good — it clearly is — but whether it's good enough to remove the human."

Translators working in higher-stakes fields argue the answer is still no. The volume of AI-translated text keeps climbing; a separate tracker, AI Productivity Hub, has flagged translation as part of a broader wave of AI tools reshaping office work. But for court filings and encyclopedia entries, the check still falls to a person.

By the numbers

• Google Translate users now translate roughly one trillion words a month, Google said in April.
• OKA has published more than 10,000 AI-assisted, human-checked articles on Wikipedia, mostly in Spanish, Portuguese and English.