Generative AI and the Reconfiguration of Skill Demand in Chinese Job Postings

The story of how we work is often told as a story of tools. For centuries, new machines have changed the nature of labor, shifting the balance between what humans do by hand and what they do with their minds. When a new technology arrives, economists and sociologists ask a specific question: does it replace the worker, or does it change the work itself? In recent years, a powerful new type of artificial intelligence has emerged, one capable of generating text, code, and analysis. This technology does not just automate simple, repetitive motions; it reaches into the complex cognitive tasks that have traditionally defined professional careers. The central question for the modern economy is whether this shift forces employers to stop hiring people for certain jobs, or if it simply asks those same people to do their jobs in a different way.

To answer this, a researcher at Shaanxi Normal University in China, Yuan Dong, turned to the digital marketplace itself. Instead of interviewing workers or tracking their wages, the study looked at the job advertisements that companies post online. These ads are a direct window into what employers are thinking at a specific moment. They reveal the tasks a company expects a new hire to perform and the qualifications they demand before an interview even begins. By analyzing millions of these postings over several years, the study sought to see if the arrival of generative artificial intelligence changed the blueprint of the jobs being offered. The researcher did not look at whether workers were actually fired or hired, but rather at how the description of the work itself was evolving as the technology spread.

The study examined nearly 5.7 million online job postings from across China, spanning the years 2018 to 2025. The researcher mapped the text of these advertisements to a fixed list of over 2,000 specific tasks, creating a consistent way to measure what was being asked of workers over time. The analysis focused on four distinct ways employers might change their demands. First, they looked to see if the need for routine, repetitive tasks was disappearing. Second, they checked if the mix of tasks was being redistributed, meaning some tasks became more important while others faded. Third, they measured how the combination of tasks within a single job posting was changing, asking if employers were bundling different skills together in new ways. Finally, they examined the entry requirements, looking at whether companies were asking for more years of schooling or more years of work experience.

The results offered a nuanced picture that defied a simple story of total replacement. When looking at the demand for routine tasks, the study found no clear, consistent evidence that employers were suddenly asking for less of this work on average. The data did not show a massive drop in the number of routine tasks listed in job ads, nor did it show a massive increase. This suggests that for many roles, the basic structure of the work remained stable, even as the technology became available. However, the story changed when the researcher looked at how tasks were being rearranged. The study found that the distribution of tasks was indeed shifting. In regions and industries where the technology was more likely to be used, the balance of tasks changed, with some activities becoming more prominent and others less so. This indicates that while the total volume of routine work might not have vanished, the specific mix of activities within a job was being reconfigured.

Perhaps the most striking finding concerned the way tasks were bundled together within a single job posting. The study measured the "distance" between the tasks listed in a new job ad and the tasks listed in similar jobs from the previous year. The data showed that this distance increased significantly. In plain terms, the combination of skills and duties required for a specific job title was becoming more different from what was required just a year prior. Employers were not just asking for the same old set of skills; they were asking for new combinations of them. This suggests that the role of the worker is being redefined, with tasks being rearranged into new packages that likely require different ways of thinking and working. The study noted that this change was robust enough to be considered a real shift in employer demand, rather than just random noise in the data.

When it came to the qualifications employers demanded, the findings were mixed and cautious. The study did not find a consistent rise in the number of years of schooling required for jobs. Employers did not seem to be uniformly demanding higher degrees across the board. However, there was a noticeable increase in the amount of work experience requested in the detailed analysis of specific tasks. This suggests that while the formal education bar did not necessarily move higher, employers may be placing more value on practical, on-the-job knowledge as they navigate this new technological landscape. The study also highlighted that these job advertisements represent what employers hope for, not necessarily what happens in reality. A posting might ask for a specific mix of skills, but that does not guarantee the person hired will have them, nor does it prove that the company has already retrained its existing staff.

The research concludes that the impact of generative artificial intelligence is not a single, sweeping wave that wipes out old jobs or creates entirely new ones overnight. Instead, it is a process of reconfiguration. The technology is prompting employers to rethink how they assemble the duties of a job, leading to new combinations of tasks and a greater emphasis on experience. The study explicitly avoids claiming that this proves workers are losing their jobs or that they are all being forced to go back to school. Rather, it points to a shift in the incentives for workers: the value of being able to adapt and combine different skills is rising. The evidence suggests that the labor market is responding to the technology by changing the shape of the work itself, creating a demand for workers who can navigate these new, more complex task combinations. The study serves as a map of these changing demands, showing that while the core of many jobs remains, the way those jobs are put together is evolving in response to the new tools available.