The effect of LLMs for Translation: What are people using for translation in 2026?
More people are translating text than at any point in history, yet the standalone websites built for that purpose are quietly losing visitors. The trend lines look contradictory at first glance. Traditional translation tools show declining web traffic while the sheer volume of translation activity has clearly increased.
The demand did not shrink. It consolidated onto platforms with built-in distribution, and general-purpose chatbots absorbed much of the casual lookup traffic that once flowed to dedicated translation sites. People did not stop translating, they simply started doing it wherever they already were.
The chatbot curves
ChatGPT offers the clearest trajectory of the group. In November 2023 it counted 100 million weekly active users. By August 2024 it passed 200 million, then 400 million in February 2025, 800 million in October 2025, and 900 million weekly actives confirmed by OpenAI in February 2026. Separately, Sensor Tower estimates reported by Reuters put the ChatGPT app near 1 billion monthly actives in June 2026. Each milestone arrived faster than the one before it, which is the kind of curve you normally only see in slide decks drawn by people selling something.
The Gemini app tells a similar story on a shorter timeline. It went from 350 million monthly actives (a figure that surfaced in a court hearing in April 2025) to over 900 million by Google I/O in May 2026, roughly 2.5x growth in 13 months. That is the fastest curve in this dataset.
Claude is missing from the chart, and leaving it off was the only honest option. Anthropic publishes no user figures, so there is no official anchor point to plot. The third-party estimates that circulate conflict wildly: some trackers report about 30 million consumer monthly actives, while Sensor Tower counts roughly 245 million across all surfaces. Its usage also skews toward API and enterprise customers, where consumer trackers cannot see it. When a product lives mostly inside other applications, public usage charts miss most of the picture.
Google Translate doubled anyway
Google Translate had 200 million monthly users in 2012 and passed 500 million by 2016. At its 20th anniversary in April 2026, Google reported more than 1 billion monthly users across Translate, Search, Lens, and Circle to Search, with 249 supported languages. In other words, it roughly doubled right through the chatbot era, which is not what you would expect from the incumbent that chatbots were supposedly disrupting.
About a third of mobile app users now use it for active language learning rather than one-off lookups, which says something about how the product's role has changed. A tool that once answered a quick question is now part of how people study.
The explanation for its growth is simple once stated. Google Translate is embedded in Search, Chrome, Lens, and Android, so it never depended on anyone deciding to visit a website. When translation is baked into the platforms people already use, the service grows with the ecosystem instead of competing for attention on its own. This is distribution as infrastructure, the kind that becomes invisible because it is everywhere.
Try Fink for your next translation
See the translation, every change, and the why on one screen.
Open AppThe squeeze on standalone translation sites
The format that finds itself under pressure is the standalone consumer translation website, the kind of destination site that people used to bookmark and visit whenever they needed to translate a phrase or a paragraph.
DeepL offers the clearest window into this shift. Semrush web visit estimates place DeepL at 204.55 million visits in June 2025, falling to 145.43 million in June 2026, a decline of about 29 percent year over year. The June to June comparison is the cleanest one available because it removes seasonality from the picture. Even so, the monthly series wobbles enough that no single month should be over-read: a March 2026 reading near 190 million looks like an outlier, and third-party traffic panels can swing by 20 percent month over month on sampling noise alone.
Crucially, DeepL the company is doing fine. It passed 250,000 business customers in the first quarter of 2026 and has kept shipping business and voice products. The erosion is specifically in consumer web traffic, not in the enterprise side. This looks like usage drifting away from browser bookmarks, not a decline in demand for translation itself.
Reverso is the clearest case of genuine substitution. It reported 96 million monthly users in 2019 and reports 70 million today. Its core use case, checking how a phrase is actually used in context, happens to be precisely the kind of question people now type into a chatbot instead of a specialized dictionary site. The service was optimized for a workflow that large language models now handle in passing.
As a counterpoint, Naver's Papago grew straight through the chatbot era on the strength of its Korean and East Asian coverage, from 10 million monthly active users in 2019 to over 20 million in 2024. Regional depth, it turns out, can withstand the wave.
What this actually means
The thesis here is not that large language models killed translation tools as a category. It is that translation consolidated onto distribution. People are translating more text than at any point in history, they simply do it wherever they already happen to be: in search results, in their browser, in a chat window. The losers in this arrangement are the destination websites built for quick throwaway lookups. The old habit of opening a separate tab, navigating to a translation site, and pasting text has largely faded for casual use, because why bother when the answer appears right where you already are?
Which leaves an interesting group behind. The people who still deliberately open a dedicated translation tool tend to be the ones who care about the result: professionals working on contracts or documentation, language learners who want to understand why a phrase works, anyone whose translation carries real stakes. Casual lookups dispersed into the platforms people already use, while serious translation work concentrates in purpose-built tools. That group, incidentally, is who Fink Translate is built for: a translator that shows what changed in a translation and why, as explained in how Fink works.
A note on the numbers
The chart's y-axis mixes metrics that are not directly comparable, so read any visual comparison loosely. Weekly active users for ChatGPT sit alongside monthly active users for Gemini, Papago, and Google Translate, and the Google Translate figures exist only at sparse milestone announcements, so the line connecting them is interpolation rather than measurement. The DeepL series consists of panel-based web visit estimates from Semrush, which are built on sampling, not server logs.
The individual figures are anchored to primary sources wherever possible. OpenAI's announcements provide the ChatGPT trajectory, compiled by Backlinko with citations, and Reuters reporting on Sensor Tower supplies the estimate that puts the ChatGPT app near 1 billion monthly actives. The Gemini points come from Alphabet earnings calls and Google events, including the 750 million milestone announced in February 2026, plus a court filing for the April 2025 number. Google's 20th anniversary post carries the 1 billion user figure for Google Translate. The Papago series comes from Naver via The Korea Times, and Reverso's historical counts are tracked on its Wikipedia page.