Fair-treatment rules for buyers and tenants

Until 2019, a landlord buying Facebook ads could keep a rental listing away from people by age, by ZIP code, and by interest categories that stood in for disability, mobility aids among them. HUD charged Facebook with violating the Fair Housing Act in March 2019. The case closed with a June 2022 Department of Justice settlement: Meta paid $115,054, the maximum civil penalty the statute allowed, and agreed to rebuild how housing ads get delivered. The penalty was rounding-error money for Meta. The rebuild was not, and it is why your campaign dashboard greys out options that worked five years ago.

Two things carry fair-treatment liability in property marketing: who your ad reaches, and how the people who respond get handled.

The legal backbone: what actually governs this

The Fair Housing Act (FHA), US federal law from 1968, bans discrimination in the sale, rental and advertising of housing on seven protected grounds: race, colour, national origin, religion, sex, familial status and disability. HUD enforces it and publishes guidance on advertising language and on digital ad targeting (HUD's guidance on discriminatory advertising).

Two exemptions get misread constantly:

• The small owner-occupied building exemption (the "Mrs. Murphy" rule) can put a landlord outside parts of the FHA, but it does not touch the advertising prohibition. An exempt landlord who publishes a discriminatory ad is still liable for the ad.
• Housing for older persons is a genuine carve-out from familial status. A 55-plus community qualifies only if at least 80% of occupied units have one resident aged 55 or over and it publishes and verifies that policy; 62-plus communities require every occupant to meet the age. "Adults only" on a building that never did that paperwork is a violation with confidence.

Many states and cities add grounds: source of income including housing vouchers, sexual orientation, gender identity, age, immigration status. New York City protects lawful source of income, so "no Section 8" there is a direct violation of a kind that passes unnoticed in a state without that protection.

The EU frame differs. Equal Treatment Directives and national statutes such as Germany's *Allgemeines Gleichbehandlungsgesetz* (AGG) prohibit housing discrimination on grounds including ethnicity, sex, religion, disability and age, enforced across member-state bodies rather than one HUD. The AGG also allows landlords to weigh the creation of socially stable resident structures, an argument that would collapse under US disparate-impact analysis. So a screening policy cannot be copied across markets.

Whether a claim in the copy is supportable belongs to the claims lesson, and mandatory fees to the disclosures lesson (FTC's junk fees initiative). Everything below assumes that layer is handled.

Steering: the subtle violation

Steering means guiding people toward or away from a property or an area based on a protected characteristic, usually without naming it.

Phrases HUD and fair-housing testers flag repeatedly:

• "Great for young professionals" (age and familial status)
• "Near [named church]" as a leading selling point (religion)
• "Quiet, mature building" (discourages families with children)
• "Walking distance to [ethnic enclave] restaurants" used to code who lives there
• "No wheelchair access, sorry" volunteered when nobody asked

None of these name a protected class, which is what makes them expensive: they pass a first read. The test is not whether you named a class but whether a reasonable reader takes the line as a signal about who belongs here.

Targeting: exclusions, proxies and delivery

The FHA reaches how an ad is *distributed*, not only what it says. The audience logic itself, built the way the positioningpositioningThe mental space you want your brand to occupy in your target customer's mind relative to alternatives.View full definition → lesson describes, becomes the compliance object.

Both platforms that sell you the inventory also police it. Google has restricted personalised advertising for housing since 2020: no targeting by gender, age, parental status, marital status or ZIP code in the US and Canada. Meta applies the same restriction through its housing special ad category, and its Variance Reduction System, the machine-learning layer built under the 2022 settlement, adjusts who actually receives a housing ad to narrow the gap between the eligible audience and the delivered one, whatever the advertiser asked for.

The failure modes are mundane:

• The campaign is never flagged as housing, so the restrictions never engage and an illegal targeting configuration runs under clean copy.
• Exclusions are drawn by ZIP or radius for "efficiency", and the boundary tracks race or income.
• A lookalike is seeded from past buyers in a neighbourhood that was historically excluded, and the model reproduces the exclusion faithfully.
• Someone re-uses a saved audience built before the account moved to housing categories.

Disparate impact, not intent, is enough for liability, which makes the audit arithmetic rather than conscience.

Quick pre-launch targeting audit

1. Is the campaign tagged in the platform's housing category on every ad set, not just the first one?
2. Do any exclusions reachreachThe number of unique people exposed to your message in a given period. Unlike impressions, reach counts each person once, no matter how often they see it.View full definition → a protected class through a proxy: ZIP, radius, language, interest, device?
3. What seeded the lookalike, and would you defend that seed list in writing?
4. Mapped against American Community Survey data (free, at census.gov), does the geofence cut out a protected group at a rate you cannot explain commercially?

Screening: where liability moves off the ad

Marketing hands over leads, and the handover is where most complaints are born. Tens of thousands of fair-housing complaints are filed in the US each year, and disability is the most common ground, ahead of race.

Practices that create liability even with a spotless campaign:

• A blanket ban on anyone with a criminal record. HUD's 2016 guidance treats it as a disparate-impact problem because arrest and conviction rates differ sharply by race. The defensible version looks individually at what happened and when, and gives arrest records without conviction no weight at all.
• Refusing to count non-employment income. Disability benefits, pensions and child support are income; a "must be employed" rule, or a requirement that annual income reach 40 times the monthly rent, excludes disabled and older applicants at a far higher rate than anyone else.
• Occupancy limits tighter than HUD's rule of thumb of roughly two people per bedroom, which reads as a familial-status filter.
• A no-pets policy applied to assistance animals. HUD's 2020 assistance-animals notice requires the accommodation, and no pet fee or deposit attaches to it.
• Inconsistent conversation. An agent who describes the building differently to different callers is exactly what paired testing catches, and testers need not be real applicants to make a case.
• Lead forms asking family size, nationality or age before a viewing. Even if nobody reads the answer, the field is evidence that you collected it.

Third-party tenant scoring sits in the same review: buying a model does not move the liability, and the vendor's terms will say so.

What has to be on file afterwards

A complaint can reach HUD up to a year after the conduct, and a private federal suit up to two years, so the evidence exists only if someone kept it. The sign-off lesson sequences the review; the artefacts specific to fair treatment are:

1. Screenshots of the audience definition and every exclusion, per ad set, dated.
2. The written selection criteria in force that day, plus the order applications arrived and were decided.
3. Reasons recorded for each rejection, in the words used at the time.
4. Accommodation requests and what was granted.

Without those you are arguing intent against a record assembled by the complainant. Meta's outcome is the benchmark for leaders: small cheque, years of consent-decree monitoring, product rebuilt.

🎬 [VIDEO: "Fair Housing Act Advertising Rules Explained" - youtube.com/results?search_query=fair+housing+act+advertising+rules - search this term for current HUD-aligned explainer videos on compliant listing language, since specific creator content changes over time]

Key takeaways

• The FHA covers advertising and delivery, not only leasing decisions, and its advertising ban has no small-landlord exemption.
• Steering language rarely names a protected class. Audit for who the phrase implies belongs there.
• Housing special ad categories at Meta and Google strip age, gender and ZIP targeting, but only once the campaign is tagged; mistagging produces an illegal configuration under clean copy.
• Screening does the quiet damage: blanket criminal-record bans, income tests that ignore benefits, tight occupancy caps, no-pets rules applied to assistance animals.
• Keep dated audience screenshots and dated rejection reasons. Complaints arrive up to a year later at HUD, two in court.