Companies that brought generative AI into their workflows cut back on junior staff, not senior ones. Six quarters after ChatGPT arrived, junior employment at these firms had fallen about 9% relative to senior employment, mostly because they hired fewer young workers.
The paper is by Seyed M. Hosseini and Guy Lichtinger of Harvard University. Its first version is dated August 31, 2025, and it was posted on SSRN on September 8, 2025, with a revised version in June 2026.
Spotting firms that adopted AI
The authors used LinkedIn résumé and job posting data from Revelio Labs covering more than 280,000 US firms, about 65 million workers and more than 160 million employment spells from 2015 to 2025. Each position carries a standardized seniority label, so the team could track junior and senior headcount inside every firm.
To identify adopters, they looked for job ads seeking “GenAI integrators,” people hired to implement or run generative AI in a company’s work. Keyword searches flagged candidates and a large language model confirmed them. By March 2025, 10,433 firms qualified. That is only about 3.7% of firms in the sample, but because adopters are large they account for roughly 16% of employment.
Juniors fell, seniors held steady
From 2018 to 2022, junior employment moved in step at adopting and non-adopting firms. From late 2022 it dropped sharply at adopters, while senior employment showed no similar break.
The decline was concentrated in jobs most exposed to generative AI. In those occupations, junior employment fell about 7% relative to less exposed jobs over two years, while senior employment kept rising.
A separate analysis built around each firm’s own adoption date found no difference before adoption. About two quarters afterward, junior employment began to fall, reaching an 8% reduction after eight quarters.
Hiring was the main channel. Junior hires dropped substantially at adopting firms. Junior separations actually fell, and promotions rose slightly, but neither came close to offsetting the hiring slowdown.
Using a language model to map postings to standard job tasks, the authors found that tasks exposed to AI became less likely to appear in junior roles at adopting firms after 2023, a pattern not seen for senior roles. Juniors from mid-tier universities saw the largest declines.
Not rates, remote work or the tech bust
The authors tested other explanations. Adopting firms did not react more strongly to monetary policy shocks, results held when comparing jobs that could and could not be done remotely, and job postings fell at adopters, which points to weaker demand rather than a shortage of junior applicants.
Because the drop began soon after ChatGPT’s release, faster than automation usually bites, the authors suggest firms may be scaling back hiring for roles they expect AI to take over. They say their data cannot test that directly, and that if firms are overestimating AI’s near-term abilities, the decline may prove temporary.
Their adoption measure also misses “silent” adoption by employees using AI tools on their own, which likely makes the estimates conservative. The bottom rung is where the change shows: fewer junior hires at the firms building AI into their work.
Study Details:
- Title: Generative AI as Seniority-Biased Technological Change: Evidence from U.S. Résumé and Job Posting Data
- Authors: Seyed M. Hosseini, Guy Lichtinger
- Journal: SSRN working paper (Harvard University)
- Publication Date: September 8, 2025
- DOI: 10.2139/ssrn.5425555