🚨 Conclusion: As LLMs become more prevalent, understanding their nuanced effects on labor market signaling is crucial. Future hiring processes must account for these complexities to remain fair and efficient.

Using experimental data, we built a labor market model to quantify impacts. We find that LLM-driven distortions can reduce overall hiring efficiency by up to 6%, as recruiters begin discounting signals across the board due to suspicion of AI-generated content.

In our second experiment, some recruiters explicitly knew when letters were ChatGPT-assisted. We find that recruiters reward high-quality applicants who did not use AI, rating their letters higher and increasing their interview chances.

Detailed textual analysis of ChatGPT interactions revealed applicants repeatedly asked AI for help specifically on the personalized sections. Despite this effort, improvements remained minimal, showing a mismatch between demand and AI’s capabilities.

Why? Recruiters valued personalized sections— such as motivation—more than generic sections improved by ChatGPT. LLMs did not effectively enhance these critical sections.

Results: ChatGPT improved overall cover letter quality by 0.2 standard deviations. However, this improvement did not lead to increased interview invitations—meaning better letters didn’t equal better hiring outcomes.

We ran two field experiments. In the first, job-seekers wrote cover letters either with or without ChatGPT assistance. Recruiters from four global companies evaluated these letters without knowing which used LLMs.

Cover letters, essential for signaling soft skills and motivation, can now be easily written using LLMs. Our research question: Do LLM-generated cover letters help job-seekers, and do they distort hiring decisions?

The labor market is transforming rapidly due to new technologies. LLMs are very influential, with 23% of U.S. workers already using ChatGPT at their jobs.
Current literature has examined the impact of technology on work but hasn’t yet deeply explored how LLMs affect job matching

Excited to share our new working paper (with Kian Abbas Nejad, Till Wicker, and Niccolò Zaccaria) on how LLMs like ChatGPT influence labor market signals—especially cover letters—and hiring outcomes!

bit.ly/4i5q5c0