A new large-scale Stanford University study examining how AI hiring systems operate across real-world job applications is raising concerns about racial bias and the growing influence of shared hiring algorithms.
Researchers tracked 3.4 million people who submitted roughly 4 million applications across 1,700 job postings at 150 employers in 11 industries. Every application in the study was screened using the same third-party AI hiring platform.
The findings suggest that algorithmic screening tools may not only disadvantage certain racial groups, but also increase the likelihood that some candidates are repeatedly rejected across multiple employers using the same system.
Researchers Found Significant Racial Disparities
The study found evidence of racial disparities in how applicants were recommended for the next stage of hiring.
Using the Equal Employment Opportunity Commission’s “four-fifths rule” — a standard used to identify potential hiring discrimination — researchers found that 26% of Black applicants and 15% of Asian applicants applied for jobs where the AI system appeared to disadvantage their racial group.
Researchers estimated that tens of thousands more applications from Black and Asian candidates would have moved forward if those applicants had been recommended at the same rate as the most-favored group.
The study also found that bias became more visible when hiring outcomes were examined job-by-job rather than averaged across all positions. Researchers warned that broad aggregate data can mask disparities happening within specific roles or industries.
Shared Hiring Algorithms Could Create “Systemic Rejection”
Beyond racial bias, the study examined what happens when many employers rely on the same hiring technology vendor.
Researchers found that applicants who submitted multiple applications screened by the same AI system were more likely to be rejected from every position they applied for than statistical models would normally predict if employers were making independent hiring decisions.
According to the study, 10% of applicants who applied to four positions screened by the same vendor were rejected from all of them.
The researchers compared those findings against data from a separate large hiring study involving Fortune 500 firms and found that the same pattern did not appear when hiring decisions were made independently across companies.
Concerns Grow As AI Hiring Expands
AI hiring systems are now widely used across the U.S. job market, particularly for screening large volumes of applications for entry-level roles.
Researchers argued that these tools combine three factors that make them difficult to scrutinize: widespread adoption, significant influence over employment opportunities, and limited public transparency into how decisions are made.
The study’s authors say the findings highlight the need for more independent oversight and research into algorithmic hiring as AI becomes more embedded in recruiting and workforce decisions.
