Screened Out Before You Start: The AI Gatekeepers Deciding Who Gets to Rent in America

The Application You Never Really Had a Chance At

You find the listing. The rent is within reach—barely, but within reach. You fill out the application, pay the screening fee, and wait. The rejection comes back within minutes, sometimes seconds. No explanation. No appeal process. No human being on the other end. Just an automated verdict that will follow you to the next listing, and the one after that.

This is the new architecture of housing denial in America, and it is being built by a small cluster of property technology companies whose screening algorithms now process tens of millions of rental applications each year. Companies like RealPage, TransUnion SmartMove, Experian RentBureau, and a growing field of competitors have positioned themselves as the gatekeepers between prospective tenants and available housing. Their pitch to landlords and property management firms is efficiency, objectivity, and risk reduction. Their impact on low-income renters, people with past evictions, and those with any criminal history is something closer to a digital blacklist.

What the Algorithm Sees—and What It Misses

Tenant-screening algorithms draw on a range of data sources: credit scores, criminal background databases, eviction court records, income verification, and in some cases, social media signals or rental payment history compiled by third-party data brokers. The problem is not that these data points are entirely irrelevant to housing decisions. The problem is how they are weighted, combined, and applied—and how little transparency exists around any of it.

Eviction records are among the most consequential inputs, and among the most misleading. Eviction filing databases capture cases that were filed but never adjudicated—cases where a landlord filed paperwork but the tenant paid up, where the case was dismissed, or where the tenant prevailed in court. These dismissed and resolved cases appear in screening reports alongside actual eviction judgments, with no distinction made between them. A 2021 investigation by ProPublica found that millions of Americans carry eviction filing records that do not reflect any actual legal finding against them, but which nonetheless trigger automatic rejections from screening algorithms.

Criminal record data presents a parallel problem. Background check databases are notoriously error-prone, with the National Consumer Law Center documenting widespread instances of misattributed records, outdated information, and cases where charges were dropped or convictions expunged still appearing in screening reports. When an algorithm weights any criminal record as a near-automatic disqualifier, it is not engaging in risk assessment. It is perpetuating punishment indefinitely, long after the legal system has concluded its work.

Discrimination by Design

The Fair Housing Act prohibits discrimination in housing on the basis of race, color, national origin, religion, sex, familial status, and disability. It does not, on its face, prohibit algorithms. But the legal doctrine of disparate impact—affirmed by the Supreme Court in Texas Department of Housing and Community Affairs v. Inclusive Communities Project (2015)—holds that neutral-seeming policies that produce discriminatory outcomes can constitute unlawful discrimination even without discriminatory intent.

The data on who is most affected by algorithmic screening is damning. Black and Latino renters are evicted at significantly higher rates than white renters in comparable economic circumstances, a disparity documented extensively by Princeton University's Eviction Lab. Criminal records are disproportionately held by Black Americans, a direct consequence of decades of racially targeted policing and prosecution. Income instability—which many screening algorithms treat as a risk signal—falls hardest on women, single parents, and people with disabilities.

When an algorithm is trained on historical housing and eviction data, and that data reflects decades of racially discriminatory housing policy, the algorithm does not neutralize that discrimination. It encodes it, automates it, and scales it to millions of decisions per year. The bias does not disappear because a computer made the call. It accelerates.

In 2023, the Department of Justice and HUD issued guidance making clear that algorithmic tools used in housing decisions are subject to fair housing law. The Biden administration's FTC also began scrutinizing data broker practices that feed screening systems. But enforcement has been slow, and the industry has largely continued operating without meaningful accountability.

No Recourse, No Transparency, No Appeal

Perhaps the most insidious feature of the algorithmic screening ecosystem is the near-total absence of meaningful recourse for rejected applicants. The Fair Credit Reporting Act gives consumers the right to dispute inaccurate information in their credit files, but tenant screening reports often fall into regulatory gray areas that complicate enforcement. Landlords are frequently not required to tell applicants which screening service was used, what specific data triggered the rejection, or how the scoring model weighted different factors.

The screening fee—typically $25 to $75 per application—is nonrefundable regardless of outcome. For a low-income renter applying to multiple properties, those fees accumulate into a meaningful financial burden. And because the same flawed data follows an applicant from landlord to landlord, the rejection can repeat itself indefinitely, consuming time and money while the housing search drags on.

The strongest counterargument from the industry is that algorithmic screening removes the subjective, potentially discriminatory judgment of individual landlords—that a consistent, data-driven process is fairer than leaving the decision to a human being who might harbor personal biases. This argument has a kernel of validity. Individual landlord discrimination is real and well-documented.

But it mistakes the source of the problem. Replacing one form of discrimination with a faster, more scalable, and less legally visible form is not progress. It is evasion. A biased landlord can be sued. An algorithm can be relabeled as neutral risk assessment and deployed across an entire metropolitan housing market before a regulator even identifies the pattern.

The Broader Crisis This Accelerates

The algorithmic eviction problem does not exist in isolation. It is one mechanism within a housing affordability crisis that has made finding stable, affordable rental housing one of the defining economic struggles of American life in the 2020s. According to the National Low Income Housing Coalition, there is a shortage of more than seven million affordable rental units for the lowest-income Americans. In that environment of scarcity, screening tools function not merely as filters but as walls—ensuring that the people with the fewest options face the most barriers.

Residential segregation, which the Fair Housing Act was designed to dismantle, is being reconstructed through these systems in ways that are difficult to litigate and nearly impossible for the average renter to see, let alone challenge. The ZIP code you can access determines the school your children attend, the grocery stores within reach, the air quality you breathe, and the wealth you can accumulate through stable housing over time. An algorithm that keeps you out of certain buildings is not making a neutral actuarial judgment. It is making a decision about your future.

Congress should mandate full transparency in algorithmic tenant screening, require landlords to disclose the specific basis for any rejection, ban the use of dismissed or expunged records in housing decisions, and fund robust FTC and HUD enforcement. Until it does, millions of Americans will keep paying application fees for decisions that were already made.