Skip to main content

AI Real Estate Ads and Fair Housing: Where Agents Get Burned

Written by Coach Emily on . Posted in Uncategorized.

By Emily Terrell — Top Coach and Speaker at Tom Ferry International. Real estate’s leading voice on AI, systems, and social media.

AI real estate ads create fair housing risk in three places: who the ad targets, who the platform delivers it to, and what the AI-written copy says. You don’t have to mean to discriminate to be liable. This post breaks down where agents get exposed and the six-step check to run before any ad goes live.

Key Takeaways

  • You’re responsible for who your ad reaches and who it excludes, not just for what it says.
  • The most common exposure comes from “neutral” filters that act as proxies for protected classes: zip codes, interests, and lookalike audiences.
  • AI copy tools default to buyer-lifestyle language, and that language can read as steering.
  • Federal AI advertising guidance was withdrawn, but the Fair Housing Act itself didn’t change.
  • A five-minute pre-launch check removes most of the risk.

Is it legal to use AI to write and target real estate ads?

Yes. Using AI isn’t the problem. What the AI produces and what you tell it to target can be.

Here’s the thing nobody wants to tell you: the Fair Housing Act doesn’t care how the ad was made. Section 804(c) prohibits housing ads that indicate a preference based on a protected class. If a model wrote the copy and a platform picked the audience, the ad still carries your name and your license number.

The protected classes under federal law are race, color, religion, sex, disability, familial status, and national origin. Texas mirrors these, and many cities add more. Check your local ordinance before assuming the federal list is the whole list.

Where AI ads create fair housing risk

Most agents picture discrimination as something deliberate. The exposure I see in practice is almost always accidental. It comes from stacking reasonable-sounding choices.

Proxy targeting

You’d never target by race or family status. But zip codes, interests like “parenting” or “church events,” and tight radius choices can stand in for protected classes. One filter is rarely the issue. Three stacked filters usually are, because each one narrows the audience further.

Lookalike audiences built from past clients

This is the one that surprises agents. If your closed-client list skews toward one demographic, a lookalike audience reproduces that skew at scale. The platform finds more people who resemble the people you’ve already served.

This is the exact mechanism regulators went after at Facebook. HUD charged Facebook with violations of the Fair Housing Act on March 28, 2019, and in the settlement that followed in June 2022, Meta agreed to retire the Special Ad Audience tool — a lookalike modeling product that delivered ads to users who “look like” the advertiser’s source audience. That settlement is why Meta now requires housing ads to run under its housing Special Ad Category, which limits targeting options. NMP

Algorithmic delivery skew

Even when you choose broad targeting, the platform’s delivery engine optimizes toward whoever engages. Your ad can end up in front of a narrow slice of people without you choosing that.

You’re less likely to be the main target for platform delivery on its own. But “the algorithm did it” won’t cover the choices you made upstream, like the audience you uploaded or the campaign type you picked to get around housing restrictions.

AI-written ad copy

This is the risk I’d take most seriously if you’re writing listing ads in ChatGPT or Claude. Out of the box, AI writes about the buyer, not the house. That’s where steering language shows up:

  • “Perfect for young families”
  • “Ideal for empty nesters”
  • “Walking distance to St. Mary’s”
  • “Exclusive, quiet neighborhood”
  • “Great for working professionals”

Each of these indicates who the home is “for,” and that’s the problem. Describe the property. Let the buyer decide whether it fits their life.

AI-generated lifestyle images

If every generated image of a family in the backyard shows the same demographic, the image signals a preferred buyer as clearly as words would. Few agents are watching this yet. That’s the reason to get ahead of it.

What changed at the federal level, and why it doesn’t protect you

In 2024, HUD released guidance addressing the applicability of the Fair Housing Act to advertising of housing opportunities through online platforms that use targeted ads, including a recommendation to avoid targeting options that directly describe or relate to FHA-protected characteristics, or that are effectively proxies for such protected characteristics — either alone or in combination. Consumer Financial Services Law MonitorConsumer Financial Services Law Monitor

That guidance is gone. The Federal Register withdrawal notice lists an effective date of withdrawal of September 17, 2025. According to the National Apartment Association, these pre-existing guidelines have not been replaced with alternatives, leaving it to the courts to interpret the standards that must be followed. federalregisterNational Apartment Association

The truth is that less guidance doesn’t make you safer. It means you have less clarity. The statute still stands. Private plaintiffs, fair housing organizations, and state agencies can still bring claims. As a National Mortgage Professional analysis put it, the federal pullback merely redistributes risk and doesn’t eliminate it. NMP

The withdrawn guidance is still a useful checklist. You can read HUD’s original 2024 announcement for context.

The six-step check before any AI ad goes live

It’s not the “what”. It’s the actual “how” to do it. Run this before every listing ad. It takes about five minutes.

  1. Declare the ad as housing. On Meta, every ad promoting a specific property or real estate service runs under the housing Special Ad Category. Don’t run listing ads as “brand awareness” or “engagement” campaigns to unlock more targeting.
  2. Kill the lookalikes. Don’t build lookalike audiences from past-client lists for listing ads. Use broad geography.
  3. Fix the prompt, not just the output. Add this line to every listing prompt you’ve saved: “Describe the property, not the buyer. Don’t reference family status, age, religion, nationality, disability, or who the home is ‘perfect for.'”
  4. Run the house-or-person test. Read each sentence of the copy. If it describes a person instead of the property, rewrite it.
  5. Check the images. If you’re using AI-generated lifestyle images, make sure they don’t show a single type of buyer. Or skip people entirely and show the property.
  6. Screenshot your settings. Save the targeting and campaign settings for every ad. If a complaint ever comes up, documentation showing you chose broad targeting is your best evidence.

Here’s the honest tradeoff: broad targeting will probably cost you more per lead than a tight, stacked audience. That’s the price of doing this right. In my experience, compliant listing ads that describe the house well still perform.

“AI will write you a listing ad in thirty seconds. It’ll also tell buyers who the house is ‘perfect for’ in the same thirty seconds. The fix is one line in your prompt, and most agents haven’t added it.”
— Emily Terrell, Tom Ferry Coach

Who needs to take this most seriously

Every agent running ads needs this. Four situations carry more risk:

  • Agents running their own Meta ads without brokerage review. You’re the advertiser of record, and there’s no second set of eyes.
  • Teams using lookalike audiences built from their CRM. Your past clients reflect your past referral network, and that network may not reflect your market.
  • Agents who pasted AI copy straight into listings last year. Go back and audit what’s still live.
  • Anyone advertising rentals. Rental ads draw more testing activity from fair housing organizations. [Likely — confirm with your brokerage compliance team before publishing this line.]

Common mistakes

  • Treating AI output as finished copy. It’s a first draft, and a fair housing review is part of the editing.
  • Assuming the platform keeps you compliant. Special Ad Category limits some targeting options. It doesn’t review your copy or your uploaded audiences for you.
  • Believing the guidance rollback changed the law. It didn’t.
  • Writing for your ideal client in ad copy. Your ideal client belongs in your marketing strategy, not in the listing description.
  • Skipping documentation. If you can’t show what you targeted, you can’t defend it.

Frequently Asked Questions

Can real estate agents use AI to write listing ads?

Yes. No law prohibits using AI to write listing ads. The Fair Housing Act applies to the finished ad, whoever or whatever wrote it. If an AI-generated copy indicates a preference based on a protected class, such as “perfect for young families,” the agent who ran the ad is responsible. Review every AI draft before it is published.

Is “perfect for families” a fair housing violation?

It can be. Familial status is a protected class under the Fair Housing Act, and phrases that suggest a home is meant for a particular type of household can indicate a preference. The safer approach is to describe features, such as “four bedrooms” or “fenced backyard,” and let buyers decide whether the home fits their lives.

Did HUD withdraw its AI advertising guidance?

Yes. HUD withdrew its 2024 guidance on applying the Fair Housing Act to digital platform advertising, effective September 17, 2025. The withdrawal removed the guidance document only. The Fair Housing Act still applies to housing ads, and private parties and state agencies can still bring claims.

Are lookalike audiences allowed for real estate ads?

Lookalike audiences built from past-client lists carry real fair housing risk because they can reproduce the demographics of your existing clients. Meta retired its original lookalike tool for housing ads as part of a 2022 settlement. For listing ads, broad geographic targeting is the safer choice.

Do I have to use Meta’s Special Ad Category for listing ads?

Yes. Meta requires ads for housing, including property listings and real estate services, to be declared under its housing Special Ad Category, which limits certain targeting options. Running listing ads under other campaign types to access more targeting puts both your ad account and your compliance at risk.

How do I make my AI prompts fair housing safe?

Add one instruction to every saved listing prompt: describe the property, not the buyer, and don’t reference family status, age, religion, nationality, disability, or who the home is “perfect for.” Then read the output and rewrite any sentence that describes a person instead of the property.

The actual takeaway

You don’t need to stop using AI for your ads. You need one line in your prompts, one test before you publish, and one screenshot after. That’s the whole system. I post the AI workflows I actually use in my business every week. Follow along on Instagram at @coachemilyterrell.

This post is general information, not legal advice. For questions about a specific ad or your local laws, talk to your brokerage’s compliance team or a fair housing attorney.