The Impact of Generative AI on Job Opportunities for Junior Software Developers

• In Event: AI, Labor Markets, and Economic Impacts

Abstract

The labor market consequences of generative artificial intelligence (AI) remain theoretically and empirically unsettled. Emerging evidence suggests that AI increases productivity across workers within the same occupation (Cui et al. 2025, Brynjolfsson et al. 2025b, Hui et al. 2024) but depends on worker expertise and the stage of work being performed (Hou et al. 2025). Thus, a central question is not simply whether AI increases productivity, but how these productivity gains are distributed across workers and how firms translate them into hiring decisions.Software development provides an ideal setting in which to study this question for three reasons. First, the public release of large language models (LLMs), such as ChatGPT, initiated sudden and widely adopted changes to software development. Second, the capabilities of LLMs overlap closely with many tasks performed by software developers, such as generating code, explaining syntax, and testing implementation, especially at junior levels where work is more narrowly defined, routine, and easily reviewed  (Demirci et al. 2025, Handa et al. 2025). And third, most software jobs are posted online, revealing how firms update both the number as well as the experience requirements and skill contents of those positions in near real-time.We use the public release of ChatGPT in November 2022 as a natural experiment to examine how generative AI reshaped employer demand within software development. Using the near-universe of U.S. online job vacancies from Lightcast, we estimate a differences-in-differences model comparing software developer vacancies requiring three or fewer years’ experience to those requiring four or more years’ experience. We also provide triple-difference estimates to account for underlying trends affecting all computer and mathematical occupations and perform a placebo check with other STEM occupations that were less exposed to ChatGPT.We documented three main findings. First, the widespread introduction of generative AI resulted in a 16.3 percent drop in the relative proportion of junior- versus senior-level software developer job vacancies. This finding holds even when accounting for month and location fixed effects, firm size, population density, industry composition, and other labor market trends affecting computer and mathematical workers during this period. Second, the relative decline in demand for junior software developers appears concentrated among larger firms that typically adopt new technologies more rapidly and metro areas that have denser labor markets for talent acquisition. Third, this labor market adjustment also operated through rising experience and skill requirements within job titles rather than simply a shift toward more senior job titles.Our findings have broad implications for talent pipelines in the software industry and beyond. For recent software engineering graduates, the introduction of ChatGPT may lead to under- or unemployment in the short-run and lower earnings due to labor market scarring in the long-run. For software firms, the efficiency improvements from AI adoption may increase productivity in the short-run but disrupt the long-term pipeline of senior software engineers in the long-run. Finally, upskilling within occupations can lead to persistent labor market mismatches requiring short-term re-training for current workers and longer term changes to curriculum for future graduates.

Author

• Alicia Sasser Modestino, Northeastern University
  Presenting Author