News Center



Home News Center News

How is AI reshaping the employment market?

Nov 10-2025   




*This article is adapted from a piece by Zhang Dandan, Vice Dean of PKU NSD, Professor of Economics, and Li Qiang, Vice President of Zhaopin Limited.

 

Artificial-intelligence (AI) technology, particularly generative AI exemplified by large language models, is exerting a profound influence on the structural contradictions within China’s labour market. On the one hand, the market as a whole continues to display pronounced structural imbalances, as evidenced by worsening educational and occupational mismatches. On the other hand, in occupations that are highly exposed to AI, educational mismatch is showing unexpected signs of improvement.

 

Labour-Market Mismatch: An Increasingly Salient Structural Challenge

Labour-market mismatch denotes the discrepancy between a worker’s qualifications (or skills) and the requirements of his or her position. It manifests mainly in two forms: 1. Vertical mismatch: highly educated workers take jobs that demand fewer qualifications, resulting in “over-qualification”. 2. Horizontal mismatch: a worker’s field of study bears little or no relation to the job’s requirements, constituting “improper specialisation”.

Research shows that mismatch is both widespread and intensifying in China. Roughly 40%~45% of applications in our sample exhibit either vertical or horizontal mismatch. More alarmingly, even among applicants who ultimately achieve “nearly-hire status” matches, the share exhibiting mismatch has risen sharply in recent years. This suggests that the market’s screening mechanism is retaining more mismatched vacancies and job seekers. Underlying this trend is the “divergent pace” of higher-education expansion and economic restructuring: a large influx of highly educated job seekers has failed to find commensurate positions, forcing many to lower their expectations.

 

The Impact of AI: An Unexpected Mitigation of Mismatch in High-Exposure Occupations

We treat the release of ChatGPT in Q4 2022 as a landmark event in the current wave of AI disruption and compare labour-market outcomes before and after it.

Key finding: following the AI shock, mismatch in high-AI-exposure occupations—such as R&D, data analysis and content creation, has eased. Specifically, the share of “downward applications” among résumés received for these roles has fallen, and the mismatch rate among “nearly-hire status” has declined.

At the same time, these occupations have become more popular: application volumes have risen. Yet both the response rate and the positive-response rate from employers have fallen, indicating that employers—faced with a surge of applicants—have become more selective, thereby sharpening the matching process.

 

Mechanism Analysis: Why Can AI Alleviate Mismatch in Certain Occupations?

We propose three complementary channels:

1. Clearer signalling and higher recruitment efficiency: AI enables firms to describe job tasks and skill requirements more precisely. Analysis of job-advertisement texts shows that, after ChatGPT’s release, the number of explicitly mentioned tasks and skills for high-exposure vacancies increased. This enhanced “signalling mechanism” deters unqualified applicants, reduces indiscriminate applications and improves matching efficiency.

2. A systemic rise in occupational thresholds: Rather than simply replacing labour, AI reshapes job content by demanding deeper human–machine collaboration, thereby raising the specialisation and complexity required. The relative complexity of skills listed in vacancy announcements for these occupations has risen markedly. Higher thresholds force job seekers to evaluate their own suitability more carefully.

3. More granular task re-engineering: To integrate AI tools, firms have proactively re-designed positions and task descriptions, making them more targeted and specialised. This facilitates better labour matching.

 

Implications and Outlook

Our study reveals a salient dynamic: while the AI-led technological revolution creates turbulence, it may also enhance allocative efficiency by inducing firms to refine job design and talent selection, thereby alleviating long-standing structural mismatches.

Nonetheless, these positive effects are still confined to high-AI-exposure occupations. Across the wider labour market, overall educational mismatch persists, and the risks of involution among highly educated talent—together with deteriorating job quality—remain acute.

Future policy needs greater foresight. On the demand side, firms should be encouraged and supported to leverage technology for more precise person–job matching. On the supply side, the education system and vocational-training infrastructure must accelerate reform to meet AI-era demands for talent, knowledge structures and skills. The “growing pains” of labour-market transformation may be unavoidable, but proactive adaptation and guidance can maximise the technological dividend while minimising transition costs.

 

How is AI reshaping the employment market?

Nov 10-2025   




*This article is adapted from a piece by Zhang Dandan, Vice Dean of PKU NSD, Professor of Economics, and Li Qiang, Vice President of Zhaopin Limited.

 

Artificial-intelligence (AI) technology, particularly generative AI exemplified by large language models, is exerting a profound influence on the structural contradictions within China’s labour market. On the one hand, the market as a whole continues to display pronounced structural imbalances, as evidenced by worsening educational and occupational mismatches. On the other hand, in occupations that are highly exposed to AI, educational mismatch is showing unexpected signs of improvement.

 

Labour-Market Mismatch: An Increasingly Salient Structural Challenge

Labour-market mismatch denotes the discrepancy between a worker’s qualifications (or skills) and the requirements of his or her position. It manifests mainly in two forms: 1. Vertical mismatch: highly educated workers take jobs that demand fewer qualifications, resulting in “over-qualification”. 2. Horizontal mismatch: a worker’s field of study bears little or no relation to the job’s requirements, constituting “improper specialisation”.

Research shows that mismatch is both widespread and intensifying in China. Roughly 40%~45% of applications in our sample exhibit either vertical or horizontal mismatch. More alarmingly, even among applicants who ultimately achieve “nearly-hire status” matches, the share exhibiting mismatch has risen sharply in recent years. This suggests that the market’s screening mechanism is retaining more mismatched vacancies and job seekers. Underlying this trend is the “divergent pace” of higher-education expansion and economic restructuring: a large influx of highly educated job seekers has failed to find commensurate positions, forcing many to lower their expectations.

 

The Impact of AI: An Unexpected Mitigation of Mismatch in High-Exposure Occupations

We treat the release of ChatGPT in Q4 2022 as a landmark event in the current wave of AI disruption and compare labour-market outcomes before and after it.

Key finding: following the AI shock, mismatch in high-AI-exposure occupations—such as R&D, data analysis and content creation, has eased. Specifically, the share of “downward applications” among résumés received for these roles has fallen, and the mismatch rate among “nearly-hire status” has declined.

At the same time, these occupations have become more popular: application volumes have risen. Yet both the response rate and the positive-response rate from employers have fallen, indicating that employers—faced with a surge of applicants—have become more selective, thereby sharpening the matching process.

 

Mechanism Analysis: Why Can AI Alleviate Mismatch in Certain Occupations?

We propose three complementary channels:

1. Clearer signalling and higher recruitment efficiency: AI enables firms to describe job tasks and skill requirements more precisely. Analysis of job-advertisement texts shows that, after ChatGPT’s release, the number of explicitly mentioned tasks and skills for high-exposure vacancies increased. This enhanced “signalling mechanism” deters unqualified applicants, reduces indiscriminate applications and improves matching efficiency.

2. A systemic rise in occupational thresholds: Rather than simply replacing labour, AI reshapes job content by demanding deeper human–machine collaboration, thereby raising the specialisation and complexity required. The relative complexity of skills listed in vacancy announcements for these occupations has risen markedly. Higher thresholds force job seekers to evaluate their own suitability more carefully.

3. More granular task re-engineering: To integrate AI tools, firms have proactively re-designed positions and task descriptions, making them more targeted and specialised. This facilitates better labour matching.

 

Implications and Outlook

Our study reveals a salient dynamic: while the AI-led technological revolution creates turbulence, it may also enhance allocative efficiency by inducing firms to refine job design and talent selection, thereby alleviating long-standing structural mismatches.

Nonetheless, these positive effects are still confined to high-AI-exposure occupations. Across the wider labour market, overall educational mismatch persists, and the risks of involution among highly educated talent—together with deteriorating job quality—remain acute.

Future policy needs greater foresight. On the demand side, firms should be encouraged and supported to leverage technology for more precise person–job matching. On the supply side, the education system and vocational-training infrastructure must accelerate reform to meet AI-era demands for talent, knowledge structures and skills. The “growing pains” of labour-market transformation may be unavoidable, but proactive adaptation and guidance can maximise the technological dividend while minimising transition costs.