One Rejection, Fifty Jobs: The Algorithmic Monoculture Quietly Blacklisting You
Technology Economics

One Rejection, Fifty Jobs: The Algorithmic Monoculture Quietly Blacklisting You

16 min read

A friend of mine spent a Sunday afternoon applying to forty-three jobs. Same field, same seniority, different companies. By Tuesday morning she had eleven rejections sitting in her inbox, several of them timestamped within the same hour, a couple within the same minute.

Nobody read those applications. Nobody could have.

She thought she was getting forty-three independent shots. Forty-three different hiring managers, forty-three different Tuesdays, forty-three fresh chances for someone to see something in her. What she actually got was one decision, made by a machine she never saw, copied across every company that rents the same machine.

You did not get rejected from one job. You got rejected from every job wired to the same algorithm. And there’s a decent chance that “no” is going to sit on a server for the next year, ready to reject you again.

TL;DR:

What Algorithmic Monoculture Actually Is

The term comes from two computer scientists, Jon Kleinberg and Manish Raghavan, who wrote about what happens when lots of separate decision-makers all start using the same algorithm.

Here’s the plain-language version.

Imagine fifty companies hiring. In the old world, each one had its own people, its own taste, its own blind spots. You might be exactly wrong for company number three and exactly right for company number nineteen. The variety was the point. It meant your weird resume gap or your non-traditional background got a different read depending on who was reading.

Now imagine all fifty companies fire their screeners and subscribe to the same software. Same model. Same scoring. Same cutoff.

Suddenly those fifty “independent” decisions are not independent at all. They’re one decision, made once, applied fifty times.

Farmers have a word for planting the entire field with one genetically identical crop. Monoculture. It looks efficient right up until a single blight shows up, and then it doesn’t take out one plant, it takes out the whole field, because every plant has the exact same weakness.

That’s what we built for hiring. One blight, one bad score, and you’re not out of one job. You’re out of the field.

The Receipts: A Study of Four Million Applications

This is not a hunch. Researchers actually measured it.

A 2026 study out of Stanford (Bommasani and colleagues, titled Algorithmic Monocultures in Hiring) got hold of a dataset most people never see: roughly 4 million applications from about 3 million applicants, all screened by machine learning models from a single vendor. The vendor used gameplay-style online assessments to spit out a binary recommendation. Hire or don’t.

When you control the whole pipeline like that, you can ask a question that’s normally impossible to answer: if someone gets rejected by company A, what happens at company B?

The answer should make you angry.

Roughly 10% of applicants who applied to four positions were recommended for rejection from all four of them. Not one. Not most. All. And that rate is far higher than you’d get if the decisions were actually independent, which is the statistical fingerprint of monoculture. The model isn’t making four separate calls. It’s making one call, four times.

The researchers gave this a name: systemic rejection. Some people never get a single interview, not because forty employers each independently passed, but because one algorithm decided, and forty employers outsourced the decision to it.

It gets worse. The study found the rejections weren’t random. They fell harder on Black and Asian applicants, with behaviors in the games functioning as proxies for race. So the monoculture doesn’t just amplify rejection. It amplifies the SAME biased rejection across the entire market at once.

Let’s be real about the limits here. This is one vendor’s dataset, and a game-based assessment is not the only screening method out there. It does not prove every system everywhere behaves identically. What it proves is that when the monoculture exists, the cascade is measurable, large, and skewed. The mechanism is real. The only open question is how widely it’s deployed.

Which brings us to the next problem.

Same Gates, Different Doors

Here’s the thing the job boards never advertise: the market for hiring software is concentrated into a handful of vendors, and most employers use one of them.

Walk through the names. iCIMS, Oracle, Workday, SAP SuccessFactors, Greenhouse. On the consumer-facing side, LinkedIn and Indeed run the top of the funnel for a huge slice of everyone looking. The single biggest enterprise vendor sits around 10% of the market on its own, and the top ten vendors together account for roughly half of it.

Estimates vary, but a commonly cited figure is that over 90% of large employers use some form of automated screening before a human ever sees your application.

Stack those two facts together and the picture gets bleak fast.

When you “apply to 50 jobs,” you are not getting 50 independent evaluations. You’re getting funneled through maybe three or four screening engines, several of which share vendors, several of which share the same underlying logic about what a “good” candidate looks like.

Fifty doors. Three locks.

You can knock on every door in the building. If they all open with the same key, and you don’t have it, the number of doors stops mattering.

One Name, A Hundred Rejections: Mobley v. Workday

If this still sounds abstract, there’s a lawsuit that makes it painfully concrete.

A man named Derek Mobley applied for jobs. A lot of jobs. By his account, he was rejected from more than 100 positions, across different companies and different industries, that all had one thing in common: they screened applicants using Workday’s AI tools.

Mobley is over 40, Black, and lives with anxiety and depression. He sued in 2023, arguing that Workday’s algorithmic screening discriminated against applicants on the basis of age, race, and disability, in violation of federal law (Title VII, the Age Discrimination in Employment Act, and the Americans with Disabilities Act).

Now here’s the part that matters for the monoculture argument.

Workday tried to get the case thrown out by saying, essentially, “we’re just a software vendor, we don’t make the hiring decisions, the employers do.” A federal court in California didn’t buy it. The judge let the disparate-impact claims proceed and treated Workday as potentially liable as an AGENT of the employers using its product. In 2025, the case got the green light to move forward as a collective action, opening the door for other applicants who got filtered by the same system.

Read that back. A single company’s AI gets to function as the gatekeeper for a hundred-plus employers, and the legal system is now grappling with the fact that the gatekeeper, not just the individual employers, might be where the discrimination lives.

That is the monoculture thesis tested in a courtroom. One vendor. One hundred rejections. One person who only got told “no” by what looked like a hundred separate companies.

Why the “No” Sticks Around: The 366-Day Shelf Life

You might think a rejection is a one-time event. You applied, they passed, the moment is over, next time it’s a clean slate.

It is not a clean slate.

Your rejection is a stored data point, and it has a shelf life. Often around a year.

Let me break down why 366 days keeps showing up (yes, 366, a leap year’s worth of days), because it’s not one single policy, it’s a few overlapping realities that all land in the same place.

Legal retention. Employers that contract with the federal government are required by OFCCP rules to keep records on rejected applicants, including the stage and reason for rejection, for a minimum period that for many of them lands at one year. So that “no,” and the reason behind it, is not just allowed to be kept. For a big chunk of employers it’s legally required to be kept.

ATS defaults. The applicant tracking systems that store all this routinely default candidate-data retention to roughly a year. Your profile, your application history, your status, all of it sits in the database long after you’ve moved on and forgotten you ever applied.

Re-application windows. Plenty of employers configure their systems to auto-reject anyone who re-applies within a set window after a prior rejection. Apply again too soon and you can get bounced before a human is ever involved, on the strength of a decision that’s months old.

Add the “do not pursue” and rehire-eligibility flags that some systems carry, and the picture is clear.

You are not being evaluated fresh each time. You’re being evaluated against a record. And that record remembers your last “no” for about a year.

So the cascade isn’t only horizontal (across many employers at once). It’s also vertical (across time). One bad score can follow you forward, quietly, for the better part of a year.

Still with me? Good, because here’s where it all compounds.

The Compounding Trap

No single one of these mechanisms is the whole problem. It’s the way they stack. Here’s the full scope, laid out:

  1. Shared vendors mean your score travels with you to every employer using that vendor.

  2. Proxy features (the game behaviors, the writing samples, the assessment quirks) encode bias that then gets applied identically everywhere.

  3. Keyword gates filter you out if your resume doesn’t echo the exact phrasing of the job description, regardless of whether you can do the work.

  4. Resume-gap penalties ding the people who took time off for caregiving, illness, or a layoff, which is to say the people who most need the next job.

  5. One-way video interviews score your face, your voice, and your cadence with models that have well-documented bias problems.

  6. “Culture fit” assessments reward people who resemble the existing workforce, which quietly launders yesterday’s demographics into tomorrow’s hires.

  7. Knockout questions get copied between employers, so the same arbitrary disqualifier (a degree requirement, a years-of-experience floor) shows up across the whole field.

  8. Retention windows keep your rejection on file, so re-applying can trip an automatic “no.”

  9. National data-sharing between platforms and background-check vendors means the profile follows you across systems, not just within one.

  10. The volume illusion, where applying to more jobs feels like hedging your bets, when in a monoculture it can mean walking into the same closed door over and over with a different sign on it.

  11. No feedback, so you never learn which gate stopped you, which means you can’t fix the thing that’s silently torpedoing every application.

  12. No appeal, because there’s no human in the loop to appeal to, and the vendor isn’t accountable to you, you’re not even their customer.

Any one of these is survivable. Twelve of them stacked on top of each other, all pointing the same direction, is not a hiring process. It’s a filter you can’t see, can’t question, and can’t get around.

The Core: One Judgment Wearing Fifty Masks

Okay. Step back from the job boards for a second, because the actual issue here is bigger than hiring, and it’s worth naming precisely.

It would be easy to say “this is just capitalism” and stop. Monopolies bad, software companies bad, end of story. That’s the reflexive take, and it’s too blunt to be useful. Plenty of markets are consolidated without doing THIS to people.

Here’s the sharper version.

A job market was never fair on any single decision. It was survivable because of NUMBERS. Many independent evaluators meant your quirks got averaged out. The resume gap that read as a red flag to one person read as “life happened” to another. The weird career pivot that confused one screener intrigued the next. The diversity of judgment WAS the safety net. Not because any one judge was wise, but because there were enough of them, rolling enough separate dice, that no single bad read could end you.

Consolidation deleted that.

Not by making the market smaller. By collapsing all those independent judgments into a few shared backends, while keeping the storefront looking exactly the same. You still see thousands of job postings. Thousands of logos, thousands of “we’re hiring!” banners. The variety is all still there on the surface. Underneath, it’s a few engines making the same call over and over.

The illusion of a competitive job market is now the product. You think you’re getting fifty shots. You’re getting one shot, scored once, replayed fifty times.

And here’s the part that should land with anyone who actually believes in free markets: this betrays the market’s own central promise. The whole pitch of a free labor market is that many independent buyers compete for your labor, and competition protects you. Consolidation didn’t deliver more competition. It collapsed the competition into a shared backend and kept the storefront lit. That’s not a free market doing its job. That’s a free market wearing a costume.

You can see the same pattern wearing different clothes all over the place once you start looking:

  • Renting an apartment? A handful of tenant-screening and background-check vendors decide whether you’re “risky,” and one bad flag follows you from landlord to landlord.
  • Borrowing money? A few credit-scoring models gate your access to nearly everything, and the score travels.
  • Buying insurance? Shared underwriting algorithms increasingly decide your rate based on data you can’t see and can’t correct.

Same structure every time. Many doors, a few locks, no key, no appeal, no exit. The hiring version is just the one most people run face-first into.

Trapped Between Two Bad Takes

This topic attracts two lazy responses, and both are wrong.

Take one: “It’s just efficiency. Employers get thousands of applications, of course they automate. The algorithm is neutral, it’s faster and fairer than a tired recruiter.”

Except it isn’t neutral, the Stanford data and the Mobley case both show measurable skew, and “faster” doesn’t help you when fast means the same wrong answer delivered to your inbox in under a minute. Efficiency that industrializes a bias isn’t a feature.

Take two: “Just network your way around it. ATS rejection is a myth, real jobs come through referrals, stop whining and build relationships.”

Sure, referrals help, and you should absolutely use them. But telling everyone to route around the front door is an admission that the front door is broken, and it quietly hands the whole market to people who already have networks, which is to say the already-advantaged. “Just know someone” is not a labor policy.

The honest position is in neither camp.

The tools are real, they’re deployed across most of the market, their failures are correlated, and they’re almost entirely unaudited. AND no amount of individual hustle fully out-runs a gate that the whole field shares. Both things are true. The fix is partly personal and mostly structural, and pretending it’s all one or all the other is how nothing gets better.

What You Can Actually Do

No magic bullets here. Some of this helps you personally, some of it only moves if enough people push. Both matter.

Tier 1: Play the gate you’re stuck with

  1. Tailor to the vendor, not just the job. Mirror the job description’s exact language for keyword gates. It feels gross. It works.

  2. Keep a plain-text, cleanly formatted resume. Fancy templates with columns and graphics confuse parsers and can get you auto-dinged before scoring even starts.

  3. Space out your re-applications. If you got rejected, assume there’s a window. Don’t re-apply to the same employer a week later and trip an automatic “no.” Wait it out, ideally past that roughly one-year mark.

Tier 2: Go around the funnel

  1. Use referrals and warm intros aggressively. Yes, it’s unfair that this works. Use it anyway, and then help someone with less of a network do the same.

  2. Apply directly when you can. A company’s own careers page, or better, a real human’s inbox, sometimes routes around the broadest screens.

Tier 3: Pull your own data and push on the system

  1. Request your data. Depending on where you live, privacy laws (GDPR in Europe, CCPA in California, and a growing list of others) let you ask what’s stored about you and sometimes demand corrections or deletion. It’s tedious. It’s also the only way to see the file that’s following you.

  2. Know your local rules and use them. New York City’s Local Law 144 requires bias audits for automated employment decision tools. Illinois regulates AI in video interviews. The Colorado AI Act, which treats high-risk hiring AI as something that has to be governed, comes into force June 30, 2026. The EU AI Act classifies hiring AI as high-risk. These laws are young and uneven, but they’re the first cracks in the “the vendor isn’t accountable to you” wall.

  3. Vote and pressure accordingly. The structural fix (mandatory audits, transparency, a right to a human review, real liability for vendors) is a policy fight, not a resume-formatting fight. The Mobley case is doing more for applicants than any keyword trick ever will.

Be honest with yourself about which tier you’re in. Tiers 1 and 2 might get YOU through. Only tier 3 changes the thing that’s doing this to everyone.

Bottom Line

The job market looks like thousands of independent doors. It’s wired to a few switchboards.

That’s the whole story compressed. One auto-rejection is rarely one rejection, it’s a single decision replayed across every employer running the same engine. The decision is invisible, you don’t get to see it. It’s unaccountable, there’s no human to appeal to. And it’s sticky, it can sit on file for the better part of a year and reject you again before anyone reads a word you wrote.

My friend with the forty-three applications eventually got a job. Through a referral, of course. A human who already knew her work vouched for her, and the whole algorithmic apparatus she’d been throwing herself against for weeks simply got skipped.

She got lucky. She had someone to vouch. The entire design of the modern hiring funnel is a machine for making sure the people without someone to vouch never get seen at all.

The cameras went up on hiring while everyone was busy polishing their resumes. The infrastructure is already built, the screens are already shared, the data is already stored. The question is whether we keep treating this as a personal failing (apply more, network harder, fix your resume) or start treating it as what it actually is: a market that quietly stopped being a market, and is still charging admission like one.

You didn’t get rejected fifty times. You got rejected once. Somebody just made sure it counted fifty times.