More, but Worse: The Marketplace Effects of AI-Generated Job Posts (Previously, Generative AI and Labor Market Matching Efficiency)

 Generative AI makes it dramatically cheaper to produce the information that labor markets rely on for hiring---but cheaper communication need not improve matching.   We consider the effects of the introduction of a generative AI tool in a large scale field experiment on an online labor market, which lowered employers' search costs by randomly offering them AI-written first drafts of their job post.   The assistance was widely accepted and treated employers were 19 % more likely to post a job. It also saved employers time---treated employers spent 44% less time writing their job posts.   Despite the substantial increase in job posts, there was no discernible increase in matches in the treatment group.   We provide evidence that the treated job posts were more generic and less informative to jobseekers.   This combination of increased job post volume and reduced informativeness diluted signals of employer seriousness. A time-cost accounting implies jobseeker application time costs are roughly six times larger than employer time savings per treated employer. In a subsequent platform roll-out expanding the access of the AI tool, posting responses attenuated while lower hiring conditional on posting remained, suggesting that learning and increased adoption did not eliminate these frictions.