Does AI cover letter writing change what employers value?

On April 19, 2023, Freelancer.com launched the AI Bid Writer, which "automatically generates a cover letter tailored to the job description" and was available exclusively to Plus or higher membership tiers. Using eight months of data for PHP and Internet Marketing, covering 5 million cover letters submitted to over 100,000 jobs, the paper reports difference-in-differences estimates that access "increased textual alignment between cover letters and job posts and raised callback rates." The central result concerns the signal rather than the outcome. After the tool's introduction, "the correlation between cover letters' textual alignment and callbacks fell by 51%," which the authors call "consistent with what theory predicts if the AI technology reduces the signal content of cover letters." Employers "shifted toward alternative signals, including workers' prior work histories."

The excerpt's logic is that the tool makes a letter "ready for immediate submission," so the letter stops carrying the information it used to carry about a worker. The Discussion states the substitution directly: AI "disrupted one channel of signaling, cover letters," while employers "successfully shifted toward others, particularly past reputation scores, which partially substituted for the diminished informativeness of written applications." The platform makes those substitutes easy to read. Bids are ranked by a recommendation algorithm that depends on a "bidder review score," and employers see the ranked list. Hiring did not move in aggregate. The excerpt reports "the absence of aggregate effects on hiring outcomes" and suggests that "the market may have adjusted to maintain equilibrium matching rates."

Against the nearest notes, this is the freelance market seen from the application side. Does AI turn freelance work into validation instead of creation? argues that freelancers lose the paid practice that builds skill once their work becomes checking AI output. This excerpt does not measure skill. It shows that what employers can still read, once the cover letter degrades, is a record of past work. The paper also draws a contrast with Cowgill et al. (2024), who informed employers whether applicants used generative AI. Here employers "typically cannot discern" it, so the two findings "are not directly comparable." The substitute the market reaches for, reviews and rankings, is a record kept by an institution, which fits the institutional dependence described in What makes accountable judgment scarce when AI cognition is cheap?. The excerpt does not test institutions as such. A related pattern appears on the worker side in Which workplace cues survive AI mediation and which disappear?, where output is treated as evidence of work. Here employers shift weight away from the letter and toward the track record.

The excerpt does not establish how far this travels. The authors call the result "preliminary" and name two limits: it covers "the immediate aftermath of the AI tool's release," and it concerns short-term freelance jobs where reviews are "readily available and highly salient." In first-time job seekers or college admissions, they note, "alternative signals may be weaker or absent." The 51% figure is reported as a change after the tool's introduction, not as a difference-in-differences estimate, and the excerpt gives no standard errors for it. The implication, at that strength: on one platform in the months after launch, cover letters lost signal content, the market moved weight to track record, and aggregate hiring rates showed no effect. The excerpt does not show that the technology is "innocuous for market-level matching," and it leaves open whether substitution holds once AI spreads to other parts of an application.