The Effects of Generative Artificial Intelligence on Job Task: Evidence from Job Posts

The rapid advancement of generative artificial intelligence (GenAI) is reshaping the job landscape by altering firms' labor demands and task requirements across occupations. This study investigates the impact of the launch of GenAI on firms' demand for specific tasks within occupations, focusing on how ChatGPT has influenced the relative desirability of different task types. Building on the task-based approach by Acemoglu and Autor (2011), we classify tasks into routine and non-routine, cognitive and non-cognitive, as well as offshorable and non-offshorable categories. Leveraging the public release of ChatGPT as an exogenous shock, our empirical strategy applies a continuous difference-in-differences approach to estimate the task reallocation effects in job postings across firms with different levels of exposure to ChatGPT. Our findings reveal that GenAI drives firms’ demand for talents engaged in both non-routine cognitive and non-routine manual tasks. Similarly, GenAI accentuates the continued importance of routine-based tasks. In addition, our study also shows that GenAI reduces the offshorability of jobs, shifting the focus toward tasks that require local execution. This study is among the first to provide empirical evidence on how GenAI is altering job content and thus provides profound implications for policymakers and business stakeholders, prompting the need for strategic adjustments and the alignment with the evolving nature of job tasks in the (Gen)AI era.