Enterprise & Work

Job Listings Wanting ChatGPT Skills Jumped 170 Percent in Korea

Korean job postings mentioning ChatGPT, Claude and Gemini rose 170 percent year over year in October. Recruiters are urged to test judgment, not prompt fluency.

How AI should change job interviews
How AI should change job interviewskirill_makes_pics / pixabay
By Rebecca Stone5 min read

Why it matters

  • JobKorea data showed a 170 percent year-over-year increase in October recruitment postings mentioning ChatGPT, Claude and Gemini.
  • About 100 technology companies attended the Seoul Job Fair on Nov. 9; around 60 planned to recruit some 400 people.
  • The proposed AI-error work sample should take no more than 10 minutes and use an ordinary problem from the position.
  • Google planned a May pilot letting candidates on selected U.S. engineering teams use Gemini while reviewing and debugging code.
  • POSCO's September recruitment campaign replaced some traditional essay questions with descriptions of concrete activities and achievements for AI and robotics roles.

Recruitment postings mentioning tools such as ChatGPT, Claude and Gemini rose 170 percent year over year in October, according to JobKorea data reported by The Korea Times. That surge frames a practical problem for Korean employers: anyone can type a polished prompt, but the valuable employee knows when the resulting answer cannot be trusted.

The setting for that problem was on display on Nov. 9, when about 100 technology companies attended the Seoul Job Fair. Around 60 of them planned to recruit roughly 400 people. Visitors used AI-based job matching, met recruiters and explored a culture-fit test. Those tools may help people find opportunities. The decisive hiring question, however, is whether applicants can recognize when a persuasive AI recommendation is wrong.

By Rebecca Stone

Why AI familiarity is not the skill worth hiring

Knowing how to operate chatbots is becoming a marketable skill in Korea's technology sector. But a job advertisement that names a fashionable chatbot may discourage a capable applicant who learned on another platform. An alternative is emerging: assess transferable judgment rather than familiarity with a particular subscription service.

The proposed method is a short work sample built around a realistic, job-related error. Consider a young engineer interviewing for a robotics role. The candidate receives a short AI-generated maintenance recommendation containing a subtle but serious error. Can the applicant identify the faulty assumption, explain the operational consequence and propose a safer next step?

The same structure adapts across industries:

  • In a biotechnology role, ask the candidate to spot a mismatch between a test description and an automated summary.
  • In a sales role, provide a customer proposal with one incorrect figure and see how the candidate handles it.

The exercise should take no more than 10 minutes and draw on an ordinary problem from the position. Candidates could describe their reasoning rather than demonstrate knowledge of a particular tool. That would help employers find people who can apply judgment across different AI systems and work settings.

What is Google doing with Gemini in coding interviews?

American employers are already experimenting with this approach. In May, Business Insider reported that Google planned a pilot on selected U.S. engineering teams. Instead of forbidding AI during a coding interview, the proposed format would let candidates use Gemini while reading, debugging and improving existing code. Interviewers would examine their ability to validate AI output and find errors. Google described the test as a way to reflect how engineers actually work.

The pilot is a useful model, with a stated limit: its results have not yet established that the method predicts job performance.

Another American example comes from enterprise software company C3 AI, whose published hiring process includes technical coding challenges and, for some candidates, written assignments and presentations. C3 AI does not describe those assignments as AI-error tests. But its work-sample structure shows how employers can ask applicants to demonstrate a skill rather than merely list it on a resume. A Korean manufacturer could adapt that principle by asking two applicants to assess the same flawed AI-generated production schedule.

How should recruiters score the answers?

The test should reward someone who challenges the tool rather than simply agrees with it. Recruiters are advised to give each candidate the same realistic scenario, the same time allowance and a clear explanation of whether AI tools are permitted. Candidates should be asked to identify the error, explain what evidence is missing and describe who should make the final decision.

A strong candidate might say there is insufficient information to proceed. That answer can reveal more professional judgment than a polished but uncritical recommendation.

Scorecards should separate three things:

  • The quality of the candidate's solution.
  • The errors the candidate detected.
  • The clarity of the explanation.

A candidate who catches a dangerous flaw and asks for missing information should receive credit, even when the machine produced a fast and fluent draft.

What about fairness and applicants' time?

American recruiters also warn against turning skills-based hiring into an obstacle course. In a 2025 Society for Human Resource Management article, a Texas Oncology talent-sourcing leader criticized assessments that demand hours from applicants without timely decisions. Seoul recruiters can avoid that mistake. A brief work sample, followed by a short discussion of the applicant's reasoning, respects candidates' time and gives hiring managers comparable evidence.

Such tests can also improve the fairness of recruitment. A concise work-sample assessment gives an applicant who learned on another platform a chance to prove transferable competence. Hiring managers should check whether questions actually predict the work, and should explain what a successful answer looks like.

One limit deserves emphasis. An employer should not let an algorithm quietly substitute for that assessment. The culture-fit demonstration at the Seoul fair may help visitors explore preferences, but a seemingly exact personality match says little about whether someone can diagnose a technical failure, communicate with a client or respond responsibly to unexpected evidence. Jobseekers should remain free to challenge a recommendation and describe a different career path.

Is Korean hiring already moving this way?

There is a domestic precedent for moving away from purely conventional screening. POSCO's September recruitment campaign replaced some traditional essay questions with descriptions of concrete activities and achievements while recruiting for AI and robotics roles. Employers across sectors can extend that principle by evaluating how applicants deal with realistic problems, including AI-generated mistakes.

The market pressure behind that shift is measurable. The JobKorea figures show employers demanding AI-related skills at more than double last year's pace, while the Seoul Job Fair's 400 openings concentrated that demand in a single venue. The gap between listing a tool in a vacancy notice and verifying that a hire can use it responsibly is where the proposed work samples sit.

Job fairs are designed to connect companies with talent. Seoul's event can show a more useful approach to AI hiring: assess a person's capacity to work with the technology, question it and take responsibility for the result. That capability, the argument runs, will outlast any particular tool listed in this year's vacancy notices.

The piece was authored by Gleb Tsipursky, PhD, a behavioral scientist, CEO of Disaster Avoidance Experts and author of "The Psychology of AI Adoption at Work: From Resistance to Results" (Georgetown University Press, 2026).

Source: feed.koreatimes.co.kr

Share this article:

More from Rebecca Stone

Rebecca Stone

Show full bio

Correspondent covering companies, deals and strategy in artificial intelligence and machine learning at AI In Context.

728 articles

Related articles

  1. HackerRank's AI Interviewer Chakra Goes Public After 500,000 Interviews
  2. When AI Search Beats Google—and When It Quietly Fails You
  3. 92% of UK Graduate Employers Report Interview Misconduct as AI Cheating Surges
  4. Largest Study of ChatGPT Use Shows AI Becoming Everyday Tool
  5. 39% of British Singles Now Use AI in Their Dating Lives

« Previous articleNext article »