Large language models (LLMs), as a new generation of artificial intelligence (AI), are reshaping production and labor market structures. Existing studies mainly focus on aggregate effects in developed countries, with limited causal and structural evidence for China, and little analysis of mechanisms from the perspectives of skills, tasks, and firm heterogeneity. This paper provides empirical evidence from the Chinese context to inform high-quality employment and balanced labor market development in the era of generative AI.

This paper uses over 550,000 online recruitment postings from listed companies between January 2022 and March 2024, combining a BERT model with a continuous DID approach to examine the multidimensional impact of LLMs on labor demand. The results show no significant impact on total hiring in the short run, but a clear increase in the share of complementary occupations, along with higher wages, educational requirements, and experience thresholds. Mechanism testing indicates increased demand for advanced cognitive skills, agreeableness, and non-routine cognitive tasks, and reduced demand for extraversion and emotional stability; these shifts lead firms to raise hiring standards. Heterogeneity analysis shows that large, profitable, and cash-rich firms adjust requirements more strongly, potentially widening inter-firm disparities.

Based on these findings, this paper proposes four policy recommendations: Implement AI-oriented reskilling programs to strengthen advanced cognitive skills and human–AI collaboration capabilities; improve employment protection for workers affected by technological shocks by establishing risk monitoring systems and promoting skill-based hiring mechanisms; reform the education system to adapt to technological changes by integrating digital literacy and AI knowledge; and adopt differentiated firm support policies to reduce the cost of AI adoption for small and medium-sized enterprises.

This paper makes the following contributions: First, it provides causal evidence on the LLM impact in China using a continuous DID framework. Second, it extends analysis from quantity to quality of labor demand and distinguishes complementary and substitutable occupations. Third, it identifies mechanisms from both skill and task perspectives, offering micro-level evidence on labor market transformation in the era of generative AI.