Gender Bias in Hiring2025n = 361,000
LLMs over-favoured female names by 1-3pp in scoring (n~361K)
In a peer-reviewed study scoring ~361,000 resumes across five frontier LLMs, female-named candidates received hiring-probability scores 1-3 percentage points higher than identical male-named candidates - the opposite direction from retrieval-based ranking.
Primary source
Armstrong, Liu, Tessler & Caliskan, PNAS Nexus 4(3)
https://academic.oup.com/pnasnexus/article/4/3/pgaf089/8071848Published 2025 — sample n = 361,000 — compiled by JobCannon Research.
Why this stat matters
This figure belongs to the gender bias in hiringvertical of JobCannon's ongoing review of hiring, AI, and career-outcome research. We track it because it is one of the few primary-source estimates with a published sample, methodology, or legal record. Cite the original source first; this page exists to make the figure easy to find and link to.
Related stats
- LLMs preferred male names 52% vs 11% female (n>3M)Wilson & Caliskan, AIES 2024 — 2024
- Amazon scrapped AI recruiter that penalised 'women's' (2018)Reuters, Oct 2018 — 2018
- Blind auditions: +50pp female advance rate in symphony orchestras (Goldin & Rouse, AER 2000)Goldin & Rouse, American Economic Review 90(4) — 2000
- Child earnings penalty: 20-30%+ for mothers, near-zero for fathers (NBER WP 25545)Kleven, Landais & Sogaard, NBER WP 25545 — 2019
Find your career fit
JobCannon assessments map your strengths to careers using primary-source research like this one. Free, no signup required.
Take the RIASEC Test