Why do I get rejected from every job I apply to using the same hiring system?
Because many employers rely on the same handful of AI screening vendors, and those vendors reuse your score across companies. If one algorithm rates you poorly, applying to more roles that run through the identical pipeline can produce the same rejection again and again.
This is the central finding of what its authors describe as the largest real-world study of deployed hiring algorithms to date. In "Algorithmic Monocultures in Hiring," presented at ACM FAccT 2026 by Bommasani, Bana, Creel, Jurafsky, and Liang (from Stanford, Chapman, and Northeastern), researchers examined 3.4 million applicants, 4 million applications, 156 employers, and 11 sectors. They found that applicants whose applications were screened by the same AI vendor were more likely to be rejected from every position than they would have been if each employer had made a statistically independent decision. Among applicants who submitted four applications through the same system, 10% were rejected from all of them.
That 10% figure matters because it undermines a core assumption most job seekers make: that each application is a separate roll of the dice. When the same vendor scores you, it is often closer to rolling the same dice over and over. A weak signal does not get re-tested; it gets copied.
Does applying to more jobs actually improve my odds?
Only if those jobs run through genuinely different evaluation systems. Volume alone does not help when the pipeline behind the roles is identical.
The same study calculated that because a vendor's scores are reused across employers, applying to more roles through identical pipelines does not generate new evaluations. The authors found that assessment scores were reused for as long as 330 days. To push the chance of being systemically shut out below 0.1%, they estimate you would need to apply to at least 25 different positions. Under independent decision-making, roughly 10 applications would have achieved the same protection. In other words, the monoculture effect can more than double the number of applications required to reach the same level of safety, and even then, only if you are reaching different systems.
This reframes the familiar advice to "just apply to more." The number of applications is less important than the diversity of the systems evaluating them. Fifty applications that all flow into the same vendor's scoring engine may function, statistically, like a much smaller and far more correlated set of chances.
How do I spread applications across different systems?
Stop counting applications and start counting distinct evaluators. Before you hit submit, try to notice which screening or assessment platform an employer uses, and deliberately mix in roles that use others.
You can often infer the system from the application experience itself. Watch the URL you get redirected to when you click "Apply," the name on any skills assessment or personality test you are asked to complete, and the branding on automated status emails. Employers within the same industry frequently cluster around the same vendors, so applying to ten similar companies in one sector can quietly funnel you into one algorithm. Applying across different sectors, company sizes, and application flows increases the chance that a fresh model actually looks at you.
It also helps to reduce your dependence on any single pipeline. Referrals, direct outreach to hiring managers, recruiter relationships, and smaller employers that still review applications manually all route around automated screening entirely. These channels are slower and less scalable than mass applying, but they generate genuinely independent evaluations, which is exactly what the research suggests high-volume applicants lack.
Why does this hit so hard right now?
Because the job market is already forcing people toward higher application volume, which amplifies exposure to the same systems. A cooling market pushes seekers to apply more, and applying more through the same vendors compounds the monoculture problem.
According to the U.S. Bureau of Labor Statistics report The Employment Situation for June 2026, released July 2, 2026, employers added just 57,000 nonfarm jobs that month, with the unemployment rate at 4.2%. April and May were revised down by a combined 74,000 jobs. A slow-hire market like this one naturally leads candidates to widen their nets and send out more applications. The danger is doing so without diversifying the systems behind those applications, which can leave you spinning through the same rejection loop.
Could the algorithm be screening against people like me?
Possibly, and it may not show up in overall averages. The same study found disparities that appeared only when researchers looked position by position.
Applying the EEOC's "four-fifths rule," the Stanford HAI research team found clear racial disparities at the position level. In the study, 26% of Black applicants and 15% of Asian applicants applied to positions where the AI screened against their racial group. These disparities were masked when outcomes were averaged across all jobs, which means an individual candidate can be adversely affected even where aggregate numbers look balanced. If you belong to a group facing this kind of position-level effect, diversifying which systems evaluate you becomes even more important, because a single biased pipeline can quietly close doors that a different evaluator would leave open.
None of this means your qualifications do not matter. It means the machinery in between you and the hiring manager is more correlated than most people assume. The practical response is not despair or endless resubmission through the same funnel; it is a deliberate strategy of reaching different evaluators, mixing automated and human channels, and treating the diversity of your applications as seriously as their number.