The Economist: “A good cover letter marries an applicant’s CV to the demands of the job. It helps employers identify promising candidates, particularly those with an employment history that is orthogonal to their career ambitions. And it serves as a form of signalling, demonstrating that the applicant cares enough about the position to go through a laborious process, rather than simply scrawling their desired salary at the top of a résumé and mass-mailing it to every business in the area.
Or, at least, it used to. The rise of large language models has changed the dynamic. Jobseekers can now produce a perfectly targeted cover letter, touching on all an advertisement’s stated requirements, at the touch of a button. Anyone and everyone can present themselves as a careful, diligent applicant, and do so hundreds of times a day. A new paper by Anaïs Galdin of Dartmouth College and Jesse Silbert of Princeton University uses data from Freelancer.com, a jobs-listing site, to work out what this means for the labour market.
Comparing pre- and post-ChatGPT activity, two results stand out. The first is that cover letters have lengthened. In the pre-LLM era, the median one was 79 words long. (Since Freelancer.com attracts workers for one-off tasks, such letters are more to-the-point than those for full-time roles.) A few years later, post-ChatGPT, the median had risen to 104 words. In 2023 the site introduced its own AI tool, allowing users to craft a proposal without even having to leave the platform. The subset of applications written using the tool—the only ones that can be definitively labelled as AI-generated—are longer still, with a median length of 159 words, more than twice the human-written baseline…(More)”.
