Sit in any Seoul café and you’ll see it within minutes: a student at the next table has a ChatGPT window open beside their essay, and on the subway a commuter is feeding meeting notes into an AI for a summary. The numbers say this impression isn’t a mirage. Korea is now one of the countries where generative AI has worked its way fastest and deepest into everyday life. But look inside, and you’ll find a story that doesn’t reduce neatly to “everyone just uses ChatGPT.”
The short answer
- 75% of Korean adults have tried generative AI, and 61% are monthly active users — roughly double the OECD average (about 33%).
- No. 1 is ChatGPT (23.45 million monthly app users in April 2026, or 46% of smartphone users), ahead of Gemini (8.45 million) and Claude (2.41 million).
- Yet Gemini now leads ChatGPT on satisfaction. Year-over-year growth points the same way: Claude +1,148%, Gemini +1,034%, versus ChatGPT +34%.
- The top use is everyday information search (42.5%). But the longer people use AI, the more their goal shifts from curiosity toward work and study productivity.
- One in four active users pays for a subscription — remarkable against the 2–5% free-to-paid conversion rate typical of software.
Do Koreans really use AI that much?
Yes. In a survey run over ten weeks (February–April 2026) by ZDNet Korea and the research firm ConsumerInsight, covering 8,000 adults (800 a week), 75% had tried an AI service at least once. Of those, 81% had used one again within the past month — which means, across all respondents, 61% are monthly active users. Three out of five adults.
The number only feels big against a benchmark. According to OECD ICT-usage figures released in January 2026, generative AI adoption across member countries stood at about 33% in 2025 — one person in three. Korea runs at nearly double that.
Among people who have tried AI, 40% say they use it “almost every day” (including several times a day). This is no longer a tool people opened once out of curiosity; it has settled into daily habit.
📊 Generative AI use by occupation (ConsumerInsight, 2026)
| Occupation | Usage rate | Note |
|---|---|---|
| Tech roles | 69% | Highest |
| Professionals | 68% | Doctors, lawyers, accountants, etc. |
| Office/clerical | 67% | Largest group |
| Management | 65% | |
| All adults (avg) | 61% | ~2× the OECD average |
That combination — high adoption, high willingness to pay, and fast switching — is why global AI companies call Korea “the most demanding, most dynamic test bed.” Build something good and it monetizes immediately; build something mediocre and it gets dropped just as fast.
Which AI do they use? ChatGPT rules — but cracks are showing
You have to read two datasets side by side: a survey of what people say (ConsumerInsight) and panel data on what people actually open on their phones (Wiseapp·Retail).
📱 Usage in the past month — survey basis (ConsumerInsight, Feb–Apr 2026, multiple answers)
| Service | Usage rate | Note |
|---|---|---|
| ChatGPT | 58% | OpenAI · the de facto default |
| Gemini | 48% | Google · rising fast |
| A. (SKT) | 9% | SK Telecom · top homegrown |
| Perplexity | 6% | AI search specialist |
| Nano Banana | 6% | Image generation |
| Copilot | 4% | Microsoft |
| Grok | 4% | xAI |
| Canva | 4% | AI inside a design tool |
| Claude | 3% | Anthropic · rose to 6% late in the survey |
| Wrtn | 3% | Korean AI portal |
📈 Monthly active users (MAU) — app panel basis (Wiseapp·Retail, April 2026)
| App | MAU | Year-over-year growth |
|---|---|---|
| ChatGPT | 23.45 million | +34% |
| Gemini | 8.45 million | +1,034% |
| Claude | 2.41 million | +1,148% |
Why the two numbers differ
The surveys cover the same period, yet ChatGPT shows 58% in one and 46% (23.45M ÷ 51.22M) in the other — because the methods differ. The survey counts web browsers, work accounts, and PC use, while the app panel only counts smartphone app launches. Gemini, meanwhile, gets a possible lift from being preloaded on Android. Neither is wrong; you need both to see the full picture. And on direction the two agree: ChatGPT is No. 1, but the chase pace belongs to Gemini and Claude.
Why are people switching to Gemini?
On raw numbers, ChatGPT’s lead looks solid. On satisfaction and willingness to recommend, the ranking flips.
⭐ Satisfaction and recommendation by service (ConsumerInsight, 2026)
| Service | Satisfaction | Recommendation (pts) |
|---|---|---|
| Gemini | 81% | 78 |
| ChatGPT | 75% | 73 |
| Claude | 71% | 77 |
An OpenSurvey poll of 1,000 people aged 10–59 in July 2026 pointed the same way: overall satisfaction for Gemini (77.3%) beat ChatGPT (70.6%) by 6.7 percentage points — a reversal within half a year.
The reason looks like ecosystem. Two out of three Gemini main users are people who left ChatGPT for it. By contrast, 88% of ChatGPT main users say “ChatGPT was my first AI and I’ve stuck with it.” In other words, ChatGPT runs on inertia, Gemini on choice. For someone already living in Gmail, Google Docs, and Google Meet, Gemini wasn’t a new app — it was an extension of tools they already used.
Claude’s usage rate is still a modest 3%, but it jumped from 1% to 6% over the survey’s ten weeks. Praised for long documents and coding, it’s spreading among developers and professionals.
How do Koreans actually use AI?
Ask what people do with AI and the answers are surprisingly modest.
🎯 Main purpose of generative AI use (2025)
| Purpose | Share |
|---|---|
| Everyday information search | 42.5% |
| Work | 24.6% |
| Study | 16.0% |
| Hobbies | 12.7% |
That “search” tops the list matters. In Korea, AI first established itself not as a new kind of tool but as a substitute for the Naver or Google search box. In the OpenSurvey poll, 56.9% of generative AI users said “when I’m curious about something, I ask AI right away,” and 42.3% said they keep digging with follow-up questions instead of stopping at the first answer.
Search hasn’t all migrated, though. The same survey found news and everyday-life information still sit firmly with Naver, at 60%+ primary usage. Work and knowledge go to AI; restaurants, weather, and news go to Naver. Koreans now split the window by purpose.
For office workers, the use is more concrete. The areas rated most effective were document summarization and writing (31.2%) and material research (27.3%) — together 58.5%. The core isn’t flashy creativity; it’s automating office drudgery.
The longer you use it, the more the purpose shifts
In the same survey, the shift among heavy users was stark. "Simple curiosity" fell from 19% to 9%, and "personal purposes" from 54% to 42%, while "work and study productivity" nearly doubled from 18% to 35%. People open it out of wonder, then — once familiar — end up putting it to work.
ChatGPT the teacher, Gemini the expert, Claude the assistant
The most interesting part of Korean research isn’t adoption — it’s how people perceive their relationship with each AI. ConsumerInsight asked users “what is this AI to you?” and the answers split cleanly by service.
Overall, 56% of AI users said they perceive AI not as a mere tool or machine but as a relational presence — an assistant, expert, teacher, or colleague. Among heavy daily users, only 36% called it “just a tool,” below the 44% average. The more you use it, the more you personify it.
And people rarely use just one. Korean AI users had tried an average of 3.6 services, and engaged Gemini main users averaged more than four. Fitting different apps to different jobs — multi-AI use — is already the norm.
How different is it by generation?
The generation gap is wide, but every arrow points up.
👥 Generative AI use by age group
| Age group | Characteristic |
|---|---|
| Teens | 67.6% usage; +25.8%p year over year — the fastest growth |
| 20s | 95.9% have tried it — effectively everyone; No. 1 in time spent and sessions per person |
| 30s | Highest share of work-related use of any age group |
| 40s | Largest cohort among ChatGPT app users |
| 50s+ | Lower usage but rising; mostly information search |
The demographic mix also differs by app. Per Wiseapp·Retail, ChatGPT skewed toward women and people in their 40s, while Claude was 62.1% male and heaviest among people in their 20s. Gemini also leaned male and 20s. Same “AI app,” different homes: ChatGPT settled around 40-something women doing work and search, while Claude and Gemini settled around 20-something men with strong coding and tech-learning demand.
Where are the homegrown AIs?
A common assumption foreigners bring to Korea — a country built on Naver and Kakao, so its own AI must be strong — is only half right.
The homegrown names worth noting are SK Telecom’s A. (9% usage), Naver’s Clova Note, and the AI portal Wrtn (3%). That’s a different order of magnitude from ChatGPT and Gemini. Naver has been developing its own model, HyperCLOVA X, but hasn’t caught the global big tech in the general chatbot race.
Instead, Korean platforms changed strategy: rather than fight head-on with a chatbot of their own, they layer AI inside apps the whole country already uses.
The Korean solution — into KakaoTalk, not another flashy AI app
- Kakao partnered with OpenAI to bring "ChatGPT for Kakao" into KakaoTalk itself. Instead of asking people to install a new app, it dropped AI into the chat window already open.
- Naver chose to AI-ify search itself, expanding an "AI briefing" that summarizes answers atop search results.
- The result: one in seven Korean internet users now reads news through an AI chatbot every week.
For travelers and expats arriving in Korea, the practical takeaway is this: you’re more likely to meet AI not as a separate app but as a feature quietly soaked into services you already use.
What’s next — agents, and distrust
Finally, the surveys asked what people would hand over to an AI agent.
🤖 Roles people would delegate to an AI agent (top-2, multiple answers)
| Role | Response rate |
|---|---|
| Professional work productivity | 43% |
| Finance and asset management | 34% |
| Household and smart-home management | 26% |
| Travel and mobility | 25% |
| Healthcare and wellness | 23% |
| Commerce and shopping on your behalf | 22% |
People want a general-purpose assistant that even manages their calendar and wallet. Yet the same respondents named information errors (29%) and lack of trust (18%) as their biggest fears about using AI. OpenSurvey found the same tension: satisfaction runs high at 70–77%, while trust in search results sat in the 50s for both services.
To sum up
Korea’s AI landscape boils down to three sentences.
- Broad and deep. Three in five adults use it monthly, and one in four of them pays.
- The leader is solid but wobbling. ChatGPT dominates on sheer scale, but Gemini and Claude lead on satisfaction and growth.
- Purpose beats inertia. The average person tries three to four services and swaps them by task — which makes Korea one of the few markets where better performance shows up in market share almost immediately.
For anyone visiting Korea or watching the Korean market, the message of this data is clear. It isn’t that “Koreans use AI.” It’s that Koreans use AI comparatively — weighing one against another, and switching the moment something better comes along.
Sources
- ZDNet Korea · ConsumerInsight, "Generative AI Consumer Survey" (Feb–Apr 2026, 8,000 adults) — ChatGPT the teacher, Gemini the expert, Claude the assistant
- Wiseapp·Retail, "ChatGPT, Gemini, Claude app users at an all-time high" (released May 2026, April 2026 basis, sample 51.22 million) — AIMatters report
- OpenSurvey, "AI Search Trend Report, H2 2026" (July 2, 2026, 1,000 people aged 10–59) — OpenSurvey blog
- OECD, ICT usage data (released January 2026, 2025 basis)
- KISDI (Korea Information Society Development Institute), "The generative AI usage gap from an AI-inclusion perspective" — usage-purpose and age-group statistics
- Kakao, "ChatGPT for Kakao" announcement — Kakao Newsroom
※ Figures can differ by research body and method (survey vs. app panel), so sources are noted inline throughout.