Fair Housing in the Age of AI: Old Rules, New Problems

The Fair Housing Act has been the law for more than 50 years, but it was written for a world of humans making decisions about other humans. The arrival of AI in real estate — making recommendations, targeting ads, scoring leads, suggesting listings — has created a new category of fair housing problems that the existing legal framework wasn't designed to handle.

How Algorithmic Bias Enters the Picture

AI systems learn from historical data. In real estate, that historical data reflects decades — in some cases, centuries — of discriminatory housing patterns. When a machine learning model is trained on that data, it can reproduce and even amplify those patterns without anyone explicitly programming it to do so.

The most concrete example that has attracted regulatory attention is algorithmic ad targeting. Meta settled with HUD in 2019 over its advertising platform, which allowed housing advertisers to exclude users from seeing ads based on characteristics that proxied for race, religion, and national origin — things like zip code, language, and behavioral patterns. The platform didn't label these as discriminatory exclusions, and advertisers didn't always understand that's what they were doing. The algorithm just optimized toward the audience most likely to engage, and that audience reflected the advertiser's existing biases.

That case involved social media advertising, but the same dynamics can apply anywhere AI is used to decide who sees what in real estate.

What's Less Settled

The more difficult questions involve AI systems that weren't designed to discriminate but produce disparate outcomes anyway. A lead-scoring algorithm that deprioritizes inquiries from certain zip codes is facially neutral but may effectively screen out protected classes. A listing recommendation engine that surfaces different properties to users based on browsing behavior could, depending on what that behavior correlates with, be steering — one of the specific practices the Fair Housing Act prohibits.

The legal framework for addressing this is underdeveloped. The Department of Housing and Urban Development has issued guidance indicating that disparate impact can establish liability even without discriminatory intent, but the specific standards for AI systems remain ambiguous.

What Agents Should Know

For individual agents, the immediate practical concern is less about the algorithms behind major platforms and more about the tools they're using directly — particularly AI-powered ad targeting. If you're using a platform that lets you define your audience for a listing promotion, understanding what characteristics that audience definition includes (and excludes) is a basic due diligence step. "The algorithm decided" is not a fair housing defense.

The industry needs clearer standards here, and it doesn't have them yet. That gap represents both a legal risk and, for the brokerages and proptech companies willing to address it proactively, an opportunity to set a higher bar.

- Jason