Most fair-housing complaints don't come from agents trying to discriminate. They come from agents who chose "cozy" instead of "compact," or wrote "perfect for families" without realizing those four words are a violation of the Fair Housing Act under the way the courts have read it for thirty-five years. This is a working guide to the language patterns regulators have flagged, the legal standard courts actually apply, and a phrase-by-phrase replacement playbook for listing copy.
Why this is suddenly everyone's problem
Two things happened. First, listing description writing got automated. Tools like ChatGPT confidently produce "nestled in a vibrant walkable neighborhood, perfect for families" copy because that's what they were trained on — listings from before fair-housing language audits became common. Second, HUD and state civil rights divisions have shifted enforcement toward the language of online listings, not just the historical pattern of discriminatory rentals. A New Jersey Division on Civil Rights investigation into an automated listing platform in 2023 made the point publicly: the platform is responsible for the words it publishes, even if the agent typed them.
That puts the agent, the brokerage, and the MLS all on the hook. A complaint over a single listing description is, on average, a four- to six-figure settlement (legal fees, education requirements, public consent decree). That's the math behind what looks like cautious word choice.
The protected classes (federal floor; your state may add more)
The Fair Housing Act, 42 U.S.C. §§ 3601-3619, prohibits discrimination in the sale or rental of housing on the basis of these classes:
- Race and color
- National origin
- Religion
- Sex (and, per HUD's 2021 directive, sexual orientation and gender identity)
- Familial status (households with children under 18, pregnant women, anyone in the process of securing legal custody of a minor)
- Disability (physical or mental)
Several states add more: source of income (Section 8 vouchers), marital status, age, lawful occupation, veteran or military status, and others. Check your state's commission. The federal categories are the floor; you can't go below them, and you can be held liable under both at once.
The "ordinary reader" test — the standard the language is actually judged against
The Second Circuit's decision in Ragin v. New York Times Co. (923 F.2d 995, 2d Cir. 1991) gave us the framework that's still applied today. A listing violates §3604(c) of the FHA if it "indicates a preference, limitation, or discrimination" based on a protected class — and the test isn't what the agent intended. The test is what an ordinary reader would understand from the language.
This is the part agents miss. Intent doesn't matter. The reasonable inference of the words on the page is what matters. "Perfect for families" doesn't mean "no single people allowed" to the writer — but a court has held it indicates a preference for families, which by implication discriminates against the non-familial-status protected group. The word "perfect" carries the violation.
"Whether an ad violates §3604(c) is determined by whether the ad would indicate a preference, limitation, or discrimination to an ordinary reader, not by what the advertiser intended." — Ragin v. New York Times Co., 923 F.2d 995 (2d Cir. 1991)
The six categories of risky language, with rewrites
1. Family / household status
Anything that signals who lives in the home — including phrasing that sounds positive — risks a familial status finding. The rule isn't "no mentioning bedrooms." The rule is no language that signals who should live there.
| Avoid | Use instead |
|---|---|
| Perfect for families | 3-bedroom layout with [feature] |
| Great for kids | Fenced backyard, in-ground sprinklers, attached garage |
| Family neighborhood | Residential street, no through-traffic |
| No kids / mature buyers only | (omit entirely — this is straight discrimination) |
| Empty nesters welcome | (describe the home; don't audience-target) |
| Singles paradise | Studio with rooftop access |
2. Religion
Mentioning houses of worship by name or proximity is the most common slip. "Walking distance to St. Mary's" or "near the JCC" both signal religion. The replacement isn't to be coy — it's to describe geography without religious markers.
| Avoid | Use instead |
|---|---|
| Walking distance to church | Walking distance to [specific intersection or shopping area] |
| Near the synagogue / mosque / parish | (omit; describe streets and parks instead) |
| Christmas-light-friendly neighborhood | (omit) |
| Christian community | (do not use; describe HOA amenities if relevant) |
3. Race / national origin (including coded language)
This is where HUD has been most aggressive in the last five years. The category covers obvious slurs (vanishingly rare today) but also coded language — words that historically signaled racial composition without naming race. "Safe," "quiet," "good schools," "up-and-coming," and "traditional" have all been flagged in HUD investigations, depending on context.
| Avoid | Use instead |
|---|---|
| Quiet, safe neighborhood | Cul-de-sac, tree-lined street, low traffic |
| Good schools / great schools | [District name] school district, GreatSchools rating: X/10 |
| Up-and-coming area | Recent developments include [specific project] |
| Traditional neighborhood | Mid-century architecture, mature landscaping |
| Exclusive community | Gated community with HOA |
| English-only speakers | (never use; this is straight national-origin discrimination) |
The school question is the most-asked. You can name the school district. You can list a public rating like GreatSchools. What you can't do is editorialize ("great schools" implies a comparison that has historically tracked racial composition). The buyer evaluates the school district; you provide the facts.
4. Disability
Counterintuitively, both excluding and highlighting disability can be a problem. The exception: when describing an accessible feature factually, you may state it neutrally. The line: describe what's there, not who it's for.
| Avoid | Use instead |
|---|---|
| Perfect for wheelchairs | Step-free entry, 36-inch interior doorways |
| Not suitable for the elderly | (never use) |
| Healthy-living community | (omit; describe gym or trails neutrally) |
| Able-bodied buyers only | (never use) |
| Walking distance to [X] | [X distance] to [X] (use measurable distance instead — "walking distance" has been read as ableist in some recent rulings) |
5. Sex, sexual orientation, gender identity
HUD's 2021 enforcement directive made it explicit: §3604 protects LGBTQ+ individuals. Listings cannot signal preference (or non-preference) on these axes either.
| Avoid | Use instead |
|---|---|
| Bachelor pad | Studio with [feature] |
| Mancave / she-shed | Bonus room / detached studio |
| Master bedroom | Primary bedroom (now the MLS-preferred term in nearly every U.S. system) |
| Master closet | Primary closet / walk-in closet |
The "master bedroom" change was driven by historical association with slavery; the National Association of REALTORS® and major MLSs have endorsed "primary" since 2020. Local MLS handbooks may now require it.
6. The "cozy" problem (size-coded language)
This one surprises agents. Why is "cozy" a fair-housing risk? Because in many cases it's been read as code for "too small for a family" — disclosing what should be in the square footage. The same goes for "starter home," "doll house," and "compact charmer." None of these are direct protected-class language, but each can support a familial-status case when paired with other signals.
| Avoid | Use instead |
|---|---|
| Cozy | Compact (or the actual square footage) |
| Charming starter home | [Square feet] single-family home |
| Doll house | [Square feet] residence |
| Nestled | Located |
| Hidden gem | (omit; describe the property) |
The myths agents keep believing
"Naming a school district is fine — naming a quality rating isn't"
Half right. You can name the district. You can cite a published, third-party rating (GreatSchools, state assessment) with the source. You cannot editorialize ("great schools," "top schools") because that's your opinion of the district, and historically those opinions have tracked racial composition.
"I'm not the writer, my brokerage is"
Wrong. The agent who signed the listing agreement is named on the complaint along with the brokerage. NAR Article 10 of the Code of Ethics holds the individual REALTOR® accountable for advertising compliance, and state real estate commissions enforce against the individual license.
"It's just one word, it can't matter that much"
The Ragin standard is whether an ordinary reader would draw an inference. A single word can carry the inference if the ordinary reader would. "Walking distance to church" — six words — has been the entire complaint in published cases.
"The MLS will catch it"
Some MLSs scan for it; many don't. The few that do scan use keyword matching, which misses the coded-language patterns described above. Don't outsource compliance to the MLS — the MLS isn't on your license.
A scanning workflow that catches what you'd otherwise miss
The pattern I use on every listing I write or review:
- Draft the listing. Don't worry about compliance in this pass — write it the way you naturally would.
- Run the protected-class pass. Search the draft for each category above. The six tables in this post are a useful checklist. Replace anything that lands in the "avoid" column.
- Run the coded-language pass. Search for: cozy, charming, perfect, safe, quiet, walking distance, family, nestled, exclusive, good schools, great schools, traditional, master, bachelor, mancave, hidden gem, up-and-coming. Replace or remove.
- Run the school pass. If you mention schools, ensure it's the district name + a published rating, not your editorialization.
- Have a second person read it. Even better if they're not a real estate agent — a fresh "ordinary reader" catches what you've stopped noticing.
- Document the review. If you ever face a complaint, having the pre-publication review in your file is the most useful thing you can produce in the discovery phase.
The 30-second AI version
If you're using an AI tool to write listings, run an explicit fair-housing scan on every output before publishing. Most AI tools, including ChatGPT, will happily produce "perfect for families nestled in a quiet, safe neighborhood with great schools" — every single phrase a flag — because that's what most listing copy looked like in their training data.
The ParClark Real Estate Agent Toolkit includes a listing-description skill that runs all six category passes plus the coded-language list automatically before output. Same workflow as above, but you don't have to remember which words to scan for.
What to do if a complaint lands anyway
Three rules, in this order:
- Don't respond to the complainant. Anything you write becomes evidence. Route to your broker-of-record immediately.
- Preserve everything. The listing as it was posted, every draft, every MLS revision history entry. State commissions and HUD will ask. A missing preservation step is its own violation.
- Get counsel. Even an "I'll handle it" response from your broker-of-record should be reviewed by counsel before it goes back to the complainant. NAR Article 16 (procuring cause), state UPL rules, and the FHA mediation process all interact in ways that aren't obvious.
The ParClark Real Estate Compliance bundle includes a complaint-response skill that walks through the preservation checklist and drafts the factual reply for counsel review. The output never leaves your office until counsel signs off.
Scan your listings before you post — not after a complaint
The Real Estate Agent Toolkit and the Real Estate Compliance bundle work together. The Agent Toolkit's listing-description skill drafts listing copy with the fair-housing filter on every output. The Compliance bundle's advertising-scan + fair-housing-scan review the draft against state real estate commission rules and the patterns HUD has flagged in published cases — before you publish.