The Fair Housing Act of 1968 makes it illegal to discriminate against renters on the basis of race, colour, national origin, religion, sex, familial status, or disability. For decades, proving housing discrimination required catching a landlord making an explicitly discriminatory decision — a recorded conversation, a written policy, a pattern documented through testing. It was difficult and slow.
AI-powered tenant screening has changed the situation entirely. Landlords using algorithmic screening tools can now discriminate at scale without ever making a decision that looks discriminatory. The algorithm makes the decision. The landlord just uses the score.
How Algorithmic Tenant Screening Works
Companies like RealPage, Yardi, and AppFolio offer AI-powered tenant screening products that take an applicant’s data — credit score, rental history, income, employment status, criminal background — and produce a recommendation: approve, decline, or conditional. Landlords pay for access to these tools and follow the recommendations. They do not independently evaluate the individual applicant in most cases.
The problem is not that these systems evaluate creditworthiness. The problem is that the training data they use to build their models reflects historical patterns of discrimination. Credit scores are lower on average for Black and Hispanic renters due to decades of discriminatory lending practices that reduced homeownership rates and wealth accumulation in those communities. Criminal background data reflects decades of racially disparate policing and prosecution. Rental history data is unavailable or negative for communities with higher eviction rates, which correlates with poverty, which correlates with race.
An algorithm trained on this data will reproduce the patterns embedded in it, even without being explicitly programmed to consider race. The discrimination is baked into the inputs, not the decision rule.
The Evidence Problem
Traditional discrimination testing works by sending matched pairs of applicants — identical in every relevant characteristic except the protected class — and measuring differential treatment. You can do this with human landlords. Algorithmic systems present a different problem: the same algorithm applies the same rules to everyone. The discrimination is not in the decision process; it is in the decision inputs. Proving it requires statistical analysis of outcomes across large populations, not individual testing pairs.
HUD, the Department of Housing and Urban Development, has acknowledged the issue. It published guidance in 2022 suggesting that algorithmic screening tools could violate the Fair Housing Act under a disparate impact theory even without discriminatory intent. The guidance created a legal framework but enforcement has been minimal. The companies selling these tools have legal teams that argue their products are race-neutral by design.
RealPage and the Rent Fixing Investigation
RealPage is facing a separate but related legal problem. The Department of Justice filed a civil antitrust lawsuit in 2024 alleging that RealPage’s revenue management software — used by landlords managing millions of apartment units across the US — facilitated illegal rent price coordination. Landlords using the same software received recommended rent prices based on aggregated data from competing properties. The DOJ alleges this functioned as algorithmic price fixing.
The two cases together illustrate what happens when an industry adopts AI tools that optimise for landlord revenue: rent goes up, approvals go down for certain demographic groups, and the mechanisms are obscure enough to resist conventional legal challenge.
AI is making the rich richer at a speed history has never seen. In housing, the mechanism is not some abstract financial instrument. It is a software subscription that tells landlords which applicants to reject and what price to charge the ones they accept.