In July 2026, South Korea’s overall employment rate reached 70.3 percent — the highest reading the country has ever posted. During that same month, employment among Koreans aged 15 to 29 fell by 191,000. It was the forty-fifth consecutive monthly decline. Both numbers are official, and the space between them is exactly where Korea AI job losses have taken hold. A worker at an auto-parts plant in Gyeonggi Province put it more plainly than any statistic could: “There are no layoffs. But new hiring is definitely down.”
That sentence describes something economists have struggled to measure anywhere in the world. Companies are not firing people because of artificial intelligence. Instead, they are quietly declining to hire the next batch. Because the damage shows up as an absence rather than an event, it rarely makes headlines. Korea, however, has just produced the clearest evidence yet that the absence is real.
The Bank of Korea just put numbers on Korea AI job losses
On 18 August 2026, the Bank of Korea published an issue note with a blunt title: Is AI to Blame for the Contraction in Youth Employment? The research team, led by Oh Sam-il of the central bank’s employment research division, did something unusual. Rather than relying on surveys, they merged National Pension Service enrolment records with the national labour force survey, then tracked individual workers from June 2022 through June 2026.
The four-year picture is stark. Youth employment fell by 285,000 across the whole economy. Of that decline, 268,000 — fully 94 percent — occurred inside sectors with high exposure to AI. Meanwhile, employment among workers aged 50 and over rose by 230,000, and 173,000 of that gain landed in those very same high-exposure sectors.
In other words, the jobs did not simply disappear. Within AI-exposed industries, they changed hands across a generation.
The sector-level detail sharpens the point further. Youth employment in information services fell 31.4 percent. Publishing dropped 27.4 percent. Computer programming and systems integration lost 16.6 percent, while professional services shed 11.6 percent. These are not declining smokestack industries. Rather, they are precisely the white-collar entry points that Korean graduates have spent two decades preparing for.
One more figure captures the shift better than any of the others. The share of all new hires who were young fell from 36.9 percent to 29.3 percent in four years. The hiring did not stop. Young people simply stopped being the ones hired.
Oh Sam-il explained the mechanism in terms any reader can follow. “AI mainly handles work that follows a manual,” he said, “and that is work which can replace young people.” Senior employees, by contrast, do work that requires understanding an organisation’s character and context — the kind of judgment that resists automation.
A university degree just became a liability
Buried in the same report is the single most counterintuitive finding to come out of Korea in years. Since ChatGPT launched in November 2022, youth unemployment among four-year university graduates has averaged 7.0 percent. For young people with a junior-college education or less, the average is 5.4 percent.
To appreciate how strange that is, consider the context. South Korea sends a higher share of its young people to university than almost any other country on earth. Families spend enormous sums on private tutoring specifically to secure a four-year degree. The entire social contract assumed that the degree paid off. For the first time in the country’s modern history, that assumption has inverted — and it inverted within three years of a chatbot going public.
Notably, the reversal fits the theory rather than contradicting it. AI is strongest at codified, manual-following knowledge work. That is exactly what a fresh graduate brings to an office. Meanwhile, the trades and technical roles that junior-college graduates enter demand physical presence and tacit skill, neither of which a language model supplies.
What July 2026 looked like on the ground
The Bank of Korea report covers four years. National statistics for a single month tell the same story at higher resolution.
In July 2026, youth employment in information and communications fell by 74,517 compared with a year earlier. That is the largest monthly drop in the series since collection began in 2014. Professional, scientific and technical services lost another 23,640 young workers. Together, those two sectors accounted for 51.3 percent of the month’s entire decline in youth employment.
Youth unemployment reached 6.8 percent, the highest July reading in four years. More telling than the level is the speed: the 1.3 percentage-point jump from a year earlier was the sharpest since January 2021.
Now place that against the rest of the labour market. Employment among Koreans aged 60 and over rose by 231,000 in July, a pattern that is reshaping the whole argument about when Koreans should retire. Workers in their forties lost 31,000 positions. Overall employment grew by 108,000, which is how the country posted a record employment rate in the same month its young people recorded a forty-fifth straight decline.
A caveat belongs here, because honest numbers deserve honest handling. Korea’s youth unemployment problem long predates generative AI, and other forces are pressing on the same numbers. Manufacturing shed 68,000 jobs in July, its twenty-fifth consecutive monthly loss, while construction lost 57,000 in its twenty-seventh. Neither of those has much to do with algorithms. Korea’s labour market is absorbing a construction downturn and a manufacturing slump at the same time as whatever AI is doing. Any account that ignores this is selling something.
Korean entry-level hiring collapsed at named companies
Aggregate statistics can feel abstract. Company disclosures do not.
Researchers recently pulled the sustainability reports of eleven listed Korean IT firms with more than 1,000 employees and compared new hires in their twenties between 2023 and 2025. Across the group, those hires fell from 2,453 to 1,387 — a drop of 43.5 percent in two years.
| Company | New hires in their 20s, 2023 → 2025 |
|---|---|
| NCSoft | 665 → 254 (−61.8%) |
| LG Uplus | 280 → 96 (−65.7%) |
| SK Telecom | 218 → 108 (−50.0%) |
| Shinsegae I&C | 65 → 18 (−72.3%) |
| Naver | 20s share: 21.3% → 17.0% |
| Kakao | 20s share: 25.5% → 15.7% |
Job platforms show the same contraction from the demand side. Entry-level full-time postings at large Korean companies fell from 3,741 to 2,145 between 2024 and 2025, a decline of 43 percent. Within IT and telecommunications specifically, postings collapsed from 899 to 293 — a fall of 67 percent. Look across a longer window and the erosion deepens: total postings on the same platform dropped from 43,953 in the first half of 2024 to 22,438 in the first half of 2026.
Still, none of that proves causation on its own. For that, someone had to ask employers directly.
In late June 2026, a survey of 206 Korean companies did exactly that. Among firms that had adopted AI, 31.3 percent reported cutting hiring, while only 3.1 percent said they had expanded it. The follow-up question is the one that matters. Of the companies cutting, 60 percent cut interns and new graduates first. Managers at the 대리 level accounted for 27.5 percent, and senior staff just 12.5 percent.
The ladder is not being removed from the top. It is being removed from the bottom.
The 공채 is dying, and that matters more than it sounds
Foreign readers need one piece of context to understand why this hits Korean graduates so hard. For decades, large Korean companies hired through 공채 — a mass open recruitment round, held twice a year, that took in hundreds of new graduates at once and trained them from scratch. It was the standard entrance to corporate life, and it did not require experience.
That system is now a minority practice. According to a 2026 survey of 500 companies with 100 or more employees, 54.8 percent now use rolling, position-by-position recruitment only. Furthermore, 67.6 percent named job-relevant work experience as their single most important screening criterion. Nearly three-quarters agreed that the defining trend in Korean hiring is a shift toward role-specific recruitment.
Consider what that means for a 24-year-old with a degree and no experience. The channel built for exactly that profile is closing, and the channel replacing it screens on the one thing she cannot have yet.
There is an irony worth noting. Some 80.8 percent of Korean firms now use AI somewhere in the recruitment process itself — a jump of 46.6 percentage points in a single year. AI now sits on both sides of the desk.
Korea AI job losses may be paralysis, not replacement
Here is where the popular framing goes wrong. A more interesting explanation is available, and it changes what the fix should look like.
Noh Se-ri of the Korea Labor Institute offered the sharpest line in the entire debate: companies themselves do not know what AI can actually replace. Faced with that uncertainty, a rational employer does not fire anyone. Instead, she freezes junior hiring and brings in experienced contract workers who need no training, waiting to see how the technology settles.
If that reading is right, the story is not “AI replaced juniors.” It is closer to “uncertainty about AI froze junior hiring.” The distinction matters enormously, because the two scenarios end differently. Replacement is permanent. Paralysis eventually resolves — though possibly not before a whole cohort ages out. Under that framing, Korean entry-level hiring is not gone forever. It is simply on hold, and nobody has said for how long.
Korean workers are living inside that uncertainty right now, and the friction it creates has already become its own office phenomenon. Meanwhile, survey data suggests Korean employees have embraced AI faster than their employers have deployed it coherently, which is a strange and specific kind of organisational limbo. Companies are meanwhile bidding aggressively for senior AI specialists, which only widens the gap at the bottom.
What Korea AI job losses cost the people living through them
Statistics describe the market. They do not describe the experience.
The average job search in Korea now lasts 11.2 months. During that period, an increasing number of applicants pay for help. Cover-letter editing runs about 40,000 won. A mock interview costs 45,000 to 50,000 won per hour. Roughly 60 percent of job seekers in their twenties have used such services or seriously considered them, and one Korean newspaper headline distilled the situation into six words: If you don’t have money, you can’t even job-hunt.
Some people stop searching altogether. In 2025, 717,000 Koreans aged 20 to 39 were classified as 쉬었음 — literally “was resting.” The category is narrower than it sounds, since it excludes anyone doing childcare, housework, or formal exam preparation. It counts people who are simply not looking. That figure is up 15.3 percent from 622,000 in 2022, and it is the highest since the series began in 2003.
Elsewhere, the effect of automation is not the absence of a job but the transformation of one. Call centres offer the clearest example, because AI arrived there early. At KB Kookmin Bank, call-centre headcount fell from 1,133 to 869. For the workers who remained, the pace changed accordingly. “Seventy calls a day used to be the standard,” one agent explained. “Now I have to take 100, 120.”
Another agent, eight years into the job at a different bank, described an unexpected side effect. When the AI voice asks a customer for their name and date of birth, it sounds exactly like a voice-phishing scam. Customers panic. Then they raise their voices at the human who comes on the line next.
Is it really AI? The honest counter-argument
A responsible account has to take the strongest opposing case seriously, and there is a strong one. Not everyone accepts that Korea AI job losses are about AI at all.
In January 2026, researchers at the Economic Innovation Group published an analysis arguing that the entry-level slowdown is largely a business cycle wearing an AI costume. Their central observation is difficult to dismiss: job postings in highly AI-exposed occupations peaked in March and April of 2022 — more than six months before ChatGPT was released. That peak coincides precisely with the start of the Federal Reserve’s rate-hiking cycle. Moreover, 38 percent of workers in the most AI-exposed quintile sit in finance and IT, two of the most interest-rate-sensitive sectors in any economy.
The Dallas Fed added a scale check. Even if every American aged 22 to 25 displaced from a highly exposed occupation became unemployed, the effect on the national unemployment rate would be roughly 0.1 percentage points. Real, but not a catastrophe.
Stanford’s Digital Economy Lab, whose “Canaries in the Coal Mine” paper is the most-cited work in this field, is careful in the same direction. Its August 2026 update found no widespread displacement across the economy. What it did find is a narrow, widening gap: workers aged 22 to 25 in highly exposed occupations now sit about 19 percent below where they would be had they tracked less-exposed peers, up from 15 percent a year earlier. Crucially, the Stanford team identified the mechanism as reduced hiring rather than increased separations — the identical pattern the Bank of Korea found on the other side of the Pacific.
Britain shows a third variant. UK job postings in mid-2026 sat 11 percent below January and 32 percent below February 2020, with graduate postings at their lowest level for the time of year since 2020. Simultaneously, AI mentions hit a record 9.4 percent of all UK postings.
So what makes Korea’s evidence distinctive? Two things. First, the data source is public pension records rather than private payroll samples, which makes individual tracking far more reliable. Second, and more importantly, the Korean finding is not just about volume. It shows composition changing inside the same sectors — young workers leaving while workers over 50 arrive. That is a much harder pattern to explain with interest rates alone.
What AI job displacement in Korea means for foreigners
More than 320,000 international students are currently studying in Korea. Their situation deserves specific attention, because the visa system was built for a hiring model that is disappearing.
A 2025 survey of 805 international students found that 86.5 percent wanted to work in Korea after graduating. In practice, only 33.4 percent did. Among the 2024 graduating cohort, 36,271 students produced 4,993 hires, and 42.9 percent left the country entirely.
Visa arithmetic explains part of that gap, and the connection to hiring practices is underappreciated. An E-7-1 professional visa requires an annual salary of at least 31.12 million won. E-7-2 and E-7-3 categories require at least 25.89 million won. Those thresholds were designed for a labour market built around 공채 — stable, full-time, salaried entry roles at large companies.
As Korean employers replace mass recruitment with small, task-level, increasingly contract-based hiring, a growing share of available entry-level positions simply cannot clear those salary floors. The consequence is subtle but severe. A foreign graduate’s D-2 to E-7 conversion may fail not because the candidate was rejected, but because the job itself no longer qualifies. To be clear, no dataset currently isolates AI’s specific effect on foreign hiring in Korea; this is an inference drawn from the structural shift, and it deserves to be read as one. Anyone weighing a move should look closely at how Korea’s work visa framework maps onto the roles that actually exist.
The ladder, not the job
The Bank of Korea’s policy recommendation is quietly remarkable. Rather than proposing measures to preserve entry-level positions, the report suggests subsidising vocational training and mentoring. Read that carefully. The central bank is not arguing that the old ladder can be rebuilt. It is arguing about what should replace it.
The government appears to agree. A youth employment recovery package — still provisional in name — is expected to target training more than 200,000 young people in advanced fields by 2030 and creating roughly 300,000 jobs. Deputy Prime Minister Koo Yun-cheol said on 13 August that it would be finalised quickly. As of this writing, it has not yet been published, and no budget figure has been made public.
Whatever it contains, it faces an awkward truth. The first rung of the Korean career ladder was never really a job. It was an apprenticeship disguised as one — a place where a 24-year-old learned by doing work that, in retrospect, was not very valuable. AI is extremely good at that work. Consequently, the rung is being sawn off, and nobody has yet built the thing that teaches judgment instead.
That, ultimately, is what Korea AI job losses are really about. The country is not running out of jobs. It is running out of the cheap, forgiving, badly-paid first jobs that turned graduates into professionals.
Korea is not unique in facing this, and the OECD has been mapping the same pressures across its member economies. Korea is simply the first country whose national statistics can prove it is happening.
