Algorithmic monocultures in hiring: what the Stanford research found
Stanford Digital Economy Lab researchers studying algorithmic monoculture in hiring found that around 4% of applicants who applied to ten positions were recommended for rejection from all ten — more often than chance would predict. They also found measurable adverse impact for Asian and Black applicants. The finding matters because it means rejections are not independent events: when employers draw on similar systems, one unfavourable assessment can echo across an entire job search.
What “algorithmic monoculture” means
A monoculture is what you get when everyone plants the same crop. It is efficient, and it is fragile: a single disease can take the whole field, because there is no variation to stop it.
Applied to hiring, the term describes what happens when many employers make decisions using the same or similar algorithmic systems. Each employer believes it is making an independent judgement. Statistically, it is not — the judgements are correlated, because the underlying machinery is shared.
The core finding
The Stanford Digital Economy Lab work on algorithmic monocultures in hiring reports that:
- Roughly 4% of applicants who applied to ten positions were recommended for rejection from all ten — a rate higher than chance alone would produce.
- Adverse impact was measurable by group, with the research reporting disparities affecting Asian and Black applicants.
- The practical implication for applicants is volume: submitting more applications across more positions raises the likelihood that a candidacy reaches human review at all.
Why “higher than chance” is the important part
If ten employers assessed you truly independently, being rejected by all ten would be unlikely unless you were genuinely unsuited to all ten roles. Independence is what makes rejection informative — it tells you something real.
Once the assessments are correlated, that logic breaks. A universal rejection may reflect one systematic judgement replicated ten times rather than ten separate verdicts. The signal a job seeker naturally reads from repeated rejection — I must be unqualified — becomes unreliable.
The same qualified applicant can be filtered out repeatedly for the same underlying reason, and receive no indication that this is what happened.
What this does not say
Precision matters here, and it is worth being explicit about the limits:
- It does not say there is a single shared score that every employer can look up. There is no such score. What actually gets calculated.
- It does not say any specific vendor acted unlawfully. It is research on systemic effects, not a legal finding against a company.
- It does not establish how long any particular vendor retains data, or that a rejection at one employer is passed to another.
We mention this because inflated versions of the finding circulate widely — including specific claims about lockout periods that the research does not contain. The real result is strong enough without embellishment.
How this connects to your applicant data
If correlated systems are the mechanism, then the data those systems hold about you is the thing worth acting on. Two consequences follow:
- The vendor matters more than the employer. Sending a deletion request to one company barely touches the problem. Sending it to the vendors that supply many companies addresses the shared layer.
- Timing matters. Retained profiles, parsed resumes and assessment results persist between job searches. Clearing them is most useful before a new round of applications, not during one.
How long platforms can keep applicant data · What to do before you apply again
The practical takeaway
Repeated rejection despite being qualified is not proof that something is wrong with your application — and the research gives a concrete reason why. Applying more widely helps. So does making sure the records those systems hold about you are not carrying an old, inaccurate or unfavourable assessment into your next attempt.
Sources
- Algorithmic Monocultures in Hiring — Stanford Digital Economy Lab
Clear your applicant data before you apply again
ATS Reset writes a deletion request for your state and gives you the privacy contacts to send it to. You send it yourself, from the email address you applied with.
Keep reading
- What is an ATS score?There is no single universal score. Here is what hiring systems actually calculate, and why the shorthand is misleading.
- Can ATS data affect future job applications?How applicant records persist across employers, and where retained data can plausibly influence a later application.
- Why qualified candidates get automatically rejectedKeyword parsing, knockout questions, ranking thresholds and assessment cut-offs — the mechanics behind a fast rejection.