Latest · Ad platforms

What Meta Actually Blocks on Housing Ads, and What Doesn't

September 2, 2026

Most of what marketing teams believe about Meta and fair housing is wrong. And the wrong part is usually the part they think they already fixed.

Here's what's actually true. In 2019, Meta settled with the ACLU, the National Fair Housing Alliance, and others after HUD had already filed a formal complaint alleging that Facebook's ad platform let advertisers exclude users by race, sex, and other protected classes. Out of that settlement came the Special Ad Category, a mandatory classification for any ad related to housing, employment, or credit. If your ad is for an apartment community, it lands in this category. There is no opting out. Not even a little.

Once an ad is flagged Special Ad Category, Meta strips out a specific list of targeting options. You lose age targeting below a floor. You lose gender targeting. You lose zip code radius targeting below a certain size, and you lose a long list of "detailed targeting" interests and behaviors that Meta itself pulled from the housing category because HUD and fair housing groups flagged them as proxies for protected classes. Targeting people interested in "Christianity," or ethnicity-linked affinity segments, gone. You can't call your rep and negotiate this back. It's baked into the ad manager the moment you select Housing as your category.

Here's the part people get backwards. The restriction is on targeting, not on ad copy or images. You can still say "close to downtown" or "walkable to transit." What you can't do is build a custom audience that excludes people over 55, or layer in a lookalike audience built from a list that was itself filtered by protected class, even if you never touch the age slider yourself. HUD's guidance and the enforcement that followed made one thing clear: intent doesn't matter. If the effect of your targeting narrows delivery along a protected class line, you're exposed. It doesn't matter what you meant to do.

The myth I'd retire first: that using stock photography of only one demographic in your ad creative is a Meta-enforced fair housing violation. It isn't. Meta's ad review doesn't catch it, and the 2019 settlement doesn't touch it. That's a separate exposure, sitting under HUD's advertising guidance and under state human rights law in places like New York and Illinois, which have gone further than federal statute. Different rulebook. Different enforcement body. Real risk. Just not the one people usually point to when they bring this up.

What I'd actually do: pull your ad account's targeting settings and confirm every apartment community campaign is tagged Special Ad Category. Some ops teams set this up correctly for lease-up campaigns and then forget renewal or retention campaigns running through the same account. Those still count. Check it manually. Don't assume your agency handled it, and don't assume it carried over when someone duplicated a campaign. Duplicated campaigns have a way of duplicating the mistakes too.

What I would not do: try to route around the targeting restrictions with a third-party data broker to build a "compliant-looking" custom audience that quietly achieves the same narrow reach Meta blocked. The restriction exists because HUD and the settlement care about outcome, not mechanism. Rebuilding the same exclusion through a side door is worse than the original violation, because now there's a paper trail showing you knew, and worked around it anyway.

State law adds another layer, and it's not consistent. Some states have amended their fair housing statutes to explicitly cover digital advertising and algorithmic targeting. Others haven't touched it, which leaves federal Fair Housing Act interpretation and HUD guidance as the operative standard. If you operate across multiple states, don't assume the rules line up. They don't.

This is not legal advice. Have counsel review your ad account setup, your campaign templates, and any vendor running targeted digital advertising on your behalf before you assume you're covered.

Source: