AI job recommendation service scores evenly but drew fewer applicants

South Korea's audit board found in 2026 that the AI job recommendation service scored about 80% for top and other occupations alike. In 2022 it found lower application rates than with an older system.

  • Not rated a person

    The 2026 score comes from evaluation data, and in 2020 and 2021 jobseekers applied less often to the service's recommendations than to the older system's.

Depth runs from 1, a convenience or time saved, to 5, a life or a livelihood preserved.

Status
Partly reported benefit.
Evidence

foundstudiedreportedclaimed

An independent audit report supports the benefit.

Compared with

The older rule-based recommender on Worknet (The Work) in the same periods, from July 2020 to October 2021. Measured.

Without the system
15.22% and 18.36% of views led to applications
With it
9.45% and 12.99% of views of AI recommendations led to applications
The system’s part
In its 2022 audit the board compared application rates for the AI system's recommendations with those for an older rule-based recommender in the same periods, and the AI system's rates were lower. The 2026 hit rate measures the model on evaluation data. The 2022 report describes a DeepFM machine-learning model trained on past applications.
Not yet checked
LimitsCosts and harmsLearns from dataLater outcome

Download data Cite Suggest a correction

Benefit type
Product Personalization
System
AI 일자리 추천서비스 (AI job recommendation service), called The Work AI on Worknet
Made by
Korea Employment Information Service (한국고용정보원), for the Ministry of Employment and Labor
Used by
Korea Employment Information Service, on the Worknet jobs portal of the Ministry of Employment and Labor
Where
South Korea (national)
Stage
In routine use
Date
, when the source was published
Kind
Aggregate result
Broad benefit
Judging the Qualities of Objects, Services, or People, Providing Consultation and Advice to Others
Collections
Scrutiny and findings
Last revised
1 November 2026

Sources

  1. Audit report, by authors with no tie to the system.

    Board of Audit and Inspection of Korea (감사원). 감사보고서 - 인공지능 대비실태 Ⅰ(신뢰성 확보 분야) - [Performance audit report, AI preparedness I: ensuring reliability]. Report dated December 2025, published 5 January 2026. Chapter 3, section 4, printed p. 54 and footnotes 52 and 53.

    'AI 일자리 추천서비스'의 평가데이터상 추천직종은 총 1,097개이며 추천 상위 10개 직종 및 그 외 직종 그룹의 추천 적중률 모두 약 80% 수준임

  2. Audit report, by authors with no tie to the system. Context or contrary evidence.

    Board of Audit and Inspection of Korea (감사원). 감사보고서 - 취업알선정보망 구축 및 관리실태 - [Performance audit report, construction and management of the employment placement information network]. Report dated March 2022, released April 2022. Printed pp. 20-24 and [표 13].

    새로 개발한 AI 기반 매칭 시스템의 입사지원율 (1차 시행기간 9.45%, 2차 시행기간 12.99%)이 같은 기간 기존의 추천 시스템의 입사지원율 15.22% 및 18.36%에 비해 오히려 각각 5.77%p, 5.37%p 낮은 것으로 확인되었다.

Details

AI system

Workflow

The service compared each jobseeker's résumé with the skills that job postings require and recommended postings. A DeepFM machine-learning model trained on past applications scored candidate postings, and the service showed up to 50 of them.

Follow-up

Conditions for replication

The evaluation data must span the occupations that the service recommends, and the 2026 data covered 1,097 of them. Matching scores must track whether jobseekers apply, and in 2022 the board found almost no correlation between them.

Entry details

Sector

Public employment services (job matching).

Purpose

The Korea Employment Information Service built the AI job recommendation service to match jobseekers' résumés with the skills that job postings require and to recommend suitable postings.

30 more questions not yet checked

model, release, provider, role, configuration, inputs and outputs, autonomy, human oversight, beneficiaries, reach, distribution and access, duration, every measurement question, every money question, ownership and control, dependencies, cross-border connections, geopolitical implications, replication attempts, availability, institutional decision, grounds, action and result, success criterion, every evidence question, observed or intended, alternative assessments, related entries, other identifiers, category mappings.

Cite this record

Peter H. Vartanian. (2026). AI job recommendation service scores evenly but drew fewer applicants (AIBD-1011). Artificial Intelligence Benefits Database, edition of 1 November 2026. https://petervartanian.xyz/aibd/records/AIBD-1011.html (archived at https://doi.org/10.7910/DVN/AADSII, version 7.0)

BibTeX Data Field definitions