Claude is the second AI assistant on South Korean phones, gaining ground in downloads though not in monthly use

By Zendoric · September 26, 2026.

Claude ended August as the second most used artificial intelligence app on South Korean phones, behind ChatGPT and ahead of South Korea's SK Telecom, Perplexity and Grok. The figure comes from Mobile Index, the app traffic measurement tool from tech firm IGAWorks, reported by South Korean outlet Asia Today and translated by UPI.

In monthly active users -people who opened the app at least once during the month, on Android and iOS- ChatGPT totaled 17,099,052 against Claude's 1,499,096: more than 11 times as many. Rounding out the podium are A., the in-house assistant from South Korean carrier SK Telecom, with 1,149,684 users; Perplexity, with 603,474; and xAI's Grok, with 546,315. Claude has now held that second place for three consecutive months, since June.

In new downloads the picture is different. ChatGPT logged 647,439 installs against Claude's 286,823, a gap of little more than 2x, far narrower than the one in monthly usage. Google's Gemini was third (204,672), Grok fourth (121,605) and PolyBuzz -a chat app with AI-generated characters- fifth (67,945). Claude has now spent five straight months, since April, as the country's second most downloaded app.

There is the figure that deserves a reading of its own: the gap with ChatGPT is far narrower among those downloading the app for the first time (2.3x) than among those already using it every month (11.4x). Put another way, among new users Claude is competing on far more even ground than its installed base suggests. That is not proof it will close the gap -downloads are a noisy indicator and some of those new users will drop the app within weeks- but it is a sign that its entry traction is improving faster than its overall share, something worth tracking over the coming months before drawing firm conclusions.

As industry context, South Korea is one of the most digitalized mobile markets in the world and, unlike the United States or Europe, it has a heavyweight domestic competitor: the telecom carrier itself, SK Telecom, whose A. assistant is contesting third place with global apps such as Perplexity or Grok. It is a reminder that the consumer AI map is not drawn solely by Silicon Valley's big labs: local carriers, with direct access to their mobile customer base, are a player to be reckoned with.

For Anthropic, moreover, this data point carries extra value: the company barely publishes usage figures for Claude as a consumer app, because its business has been built mainly around the API, enterprises and Claude Code, its tool for developers. That a third party -a Korean measurement firm- offers a snapshot of its consumer traction in a demanding market like South Korea is, in practice, one of the few external gauges available of how Claude is faring outside its corporate business.

Our read is that this ranking neatly captures a disconnect we have been flagging: the one between model quality and consumer market share. According to Zendoric's own indices, which combine coding, reasoning and tool-use benchmarks, Anthropic's models score above OpenAI's in measured capability. And yet ChatGPT is more than ten times ahead of Claude in active users in South Korea. Consumer leadership is not decided only by who reasons better, but by who defined the category first and built the brand people associate by default with "asking the AI."

In the long run, that distinction between engine and distribution is good news for Zendoric's underlying thesis: the more players genuinely competing for the end user -a local carrier, a conversational search engine like Perplexity, a character app like PolyBuzz, Claude itself- the less chance there is that access to AI ends up concentrated in the hands of a single company. The abundance we champion as a horizon needs exactly that: markets where winning does not depend on a single winner, but on more and more people, in more and more countries, having several doors into the same technology.