Fair Housing AI Listing Description Checklist (Copy, Ads, Social)

In one of my classes, an agent pasted a listing description Claude had written for her and asked the room what was wrong with it. It read well. Good rhythm, real features, no typos. Then someone pointed at the last line: "a wonderful place to raise a family." Nobody had noticed, including her.

That's the problem with AI listing copy. It's fluent enough that the risky phrase slides right past you. AI tools learned from years of older listings, flyers, and ads, and plenty of that writing describes who should live in a home instead of the home itself. So the AI repeats it, confidently.

This post is a checklist you can run on any AI-written listing description, ad, or social post before it goes live. It's education, not legal advice. Your broker, your brokerage's compliance process, and your state and local rules decide what's acceptable in your market. If you want the full prompt system for writing the description in the first place, start with my AI listing description prompt, then come back here for the review.

What is the fair housing checklist for AI listing copy?

The checklist is a short set of questions you ask about every piece of listing marketing: does it describe the property or the people, does it mention anything tied to a protected class, does it make claims you can't back up, and does it follow your MLS and ad platform rules. Here's the version I teach, laid out so you can print it or keep it next to your keyboard.

Words and phrases

• Does any line say who the home is "perfect for", "ideal for", or "great for"?
• Does it mention families, kids, couples, singles, retirees, students, or any age group?
• Does it mention religion, nearby places of worship as a selling point, national origin, race, or color?
• Does it describe the neighbors, the "type" of neighborhood, safety, or crime?
• Does it make claims about school quality?
• Does it imply a physical ability is needed to enjoy the home?
• Does it use coded words your brokerage flags (check their list)?

Accuracy

• Is every feature real and confirmed by you, not invented by the AI?
• Do square footage, beds, and baths match what's entered in your MLS?

Images and video

• Do photos and video show the property, not people who signal a preferred buyer?
• Is any AI-edited or staged image disclosed the way your MLS and brokerage require?

Ads and social

• Did you choose the housing special ad category for paid social ads?
• Does your targeting avoid excluding anyone based on a protected class?

MLS remarks

• Does the copy follow your MLS rules for public remarks (character limits, what's allowed)?

Final check

• Did you read it yourself, out loud, before posting?
• Did you send anything borderline to your broker?

Who is protected under the Fair Housing Act?

The federal Fair Housing Act protects against discrimination based on race, color, religion, sex, disability, familial status, and national origin. HUD interprets "sex" to include sexual orientation and gender identity. State and local laws can add more protected classes, so the list in your market may be longer than the federal one.

Bookmark HUD's fair housing page and check it yourself instead of trusting a blog post or an AI summary, mine included. Then ask your broker which state and local classes apply where you list. That second step matters because your AI tool won't know your local ordinances unless you tell it.

In the listing copy I see in class, familial status is the class that comes up most. It's why "great for a growing family" and "adults only feel" are both problems. The first steers toward people with kids, the second steers away from them.

Why do AI tools write fair housing risky listing copy?

AI tools write risky copy because they learned from older real estate writing, and that writing is full of people-focused phrases. "Perfect starter home for young couples" was a normal line in plenty of old listings. The AI doesn't know the phrase is a problem unless you tell it.

There's a second reason. When you give the AI thin facts, it fills the gap with mood. "3 bed, 2 bath, big yard" turns into "a backyard where your kids will make memories." The fix is more property facts going in, and a separate review pass coming out.

When I teach this in class, the first thing agents ask is whether they can just add "be fair housing compliant" to the prompt. It helps. It isn't enough. A rule inside the writing prompt competes with every other instruction. A second chat with one job, finding people-focused language, catches what the first pass missed.

How do you describe the property, not the people?

You describe the property by stating features, layout, and measurable facts, and letting the buyer decide whether it fits their life. NAR's guidance on avoiding fair housing violations in advertising gives examples like "ideal for empty nesters" and "perfect for students" as risky, and "easy walk to train" as acceptable.

A simple test I give agents: could this sentence be true and useful for any buyer who walks in? "Main-floor primary suite" passes. "Great for retirees" doesn't. One describes the house. The other describes a person.

Which listing phrases are risky, and what do you write instead?

The risky phrases fall into a handful of categories: who should live there, who lives nearby, schools and safety, religion and national origin, physical ability, and exclusivity. Here's a table by category with property-based rewrites. Use it as a starting point, then add the phrases your broker flags.

A few notes on this table. "Walking distance" style phrases are a gray area: NAR's guidance treats "easy walk to train" as fine, but some brokers prefer distances because a number is accurate for everyone. On schools, many brokerages ask agents to skip quality claims entirely and point buyers to the district. Ask yours.

What about photos, video, and ad targeting?

Fair housing applies to everything in the ad, not just the words. Photos, video, and who you show the ad to all count. The words are where AI causes most of the trouble, but images and targeting are where agents forget to look.

Photos and video

• Lead with the property. Rooms, exterior, lot, and features.
• Be careful with lifestyle photos that include people. Models who all look one way, or a scene built around one kind of household, can signal a preferred buyer.
• If you use AI to stage or edit photos, follow your MLS and brokerage rules on disclosure and on what edits are allowed. Removing a power line or adding furniture can be treated very differently.
• Read your video script through the same checklist. "You and your kids will love this yard" is the same problem spoken out loud.

Paid ads and targeting

If you run listing ads on Facebook or Instagram, Meta requires you to declare the housing special ad category, and that category limits some targeting options. The specific limits have changed over time, so check Meta's Business Help Center for the current rules before you build the audience, and don't rely on what you remember from a past campaign.

The principle is the same everywhere: don't build an audience that includes or excludes people based on a protected class, directly or through a stand-in like a narrow interest that tracks closely with one group. Google and other platforms have their own housing ad policies too. Your broker may also have a review step for paid ads.

What do MLS rules say about public remarks?

Most MLSs have rules about what can go in public remarks, and they're separate from fair housing law. Common examples include character limits, limits on contact information or showing instructions in public remarks, and rules on links or agent marketing. I can't tell you your MLS's exact rules, because they vary. Read your MLS rules and regulations, or ask your broker.

Two habits help here:

1. Give the AI your exact character limit and tell it to count. AI tools are loose with counts, so paste the result back into the MLS field and see if it fits.
2. Keep private remarks private. Don't let the AI blend showing instructions or agent notes into public copy.

How do you make Claude run the fair housing check every time?

The reliable way is to put your rules and review prompt somewhere Claude reads automatically: a Project's instructions, or a skill. Then you don't have to remember to paste anything.

Option 1: a Claude Project. A Project is a workspace with its own instructions and files that every chat inside it uses. Make a "Listings" Project, paste the instructions below, and upload your brokerage's advertising guidelines if you have them. Every description, ad, and caption you draft in that Project gets the rules.

Option 2: a Claude skill. A skill is a reusable playbook Claude uses whenever it's relevant. You can build one by describing your process in a chat, and Claude packages it for you. Turn it on in Settings > Capabilities > Skills. I walk through building these in Claude skills for real estate agents.

Here's a set of Project instructions you can paste as-is:

You help me write and review real estate marketing: MLS remarks, ads,
social captions, emails, flyers, and video scripts.

Fair housing rules for everything you write:
1. Describe the property, never the people. Never say who a home is
   perfect, ideal, or great for.
2. Never mention or imply race, color, religion, sex, sexual
   orientation, gender identity, disability, familial status, national
   origin, or age. Also avoid these local protected classes: [ADD ANY
   YOUR BROKER LISTS].
3. No claims about schools, safety, crime, neighbors, or neighborhood
   character.
4. Prefer distances and facts over ability phrases ("0.5 miles to the
   station" over "walk to everything").
5. Only use features I give you. If a fact is missing, ask me.
6. Public remarks must stay under [CHARACTER LIMIT] characters and
   contain no contact info or showing instructions.

After every draft, run a fair housing review on your own draft and list
any phrase you changed or think is borderline. Mark borderline items
"ASK BROKER." This is a review aid, not legal advice.

The fair housing review prompt (copy and paste)

Use this as a separate step, in a fresh chat or as the last step of the skill. It works on any copy, including drafts from another AI tool or from a teammate.

Act as a fair housing reviewer for real estate advertising. Review the
content below. Your only job is to find language or ideas that describe,
target, or exclude people instead of describing the property.

Content type: [MLS remarks / Facebook ad / Instagram caption / email /
video script / flyer]
Platform or MLS character limit, if any: [LIMIT]

Check against this list:
1. Who-should-live-here phrases (families, kids, couples, singles,
   retirees, empty nesters, students, professionals, any age group)
2. Anything tied to race, color, religion, sex, sexual orientation,
   gender identity, disability, familial status, national origin, or
   these local classes: [LOCAL CLASSES]
3. Claims about schools, safety, crime, neighbors, or "type" of area
4. Places named in a way that could signal a preferred group
5. Phrases that assume a physical ability
6. Exclusivity or coded words: [YOUR BROKERAGE'S FLAGGED WORDS]
7. For scripts or image descriptions: scenes or people that suggest an
   ideal buyer
8. Any feature claim that is not in the facts I provided: [PASTE FACTS
   OR SAY "NO FACT SHEET"]

Return a table with columns: Exact phrase | Concern (one sentence) |
Property-based rewrite (using only facts already in the content) |
Risk level (Change / Borderline, ask broker).

Then give the full revised content with every "Change" item fixed and
"Borderline" items left as-is but marked [ASK BROKER].

If you find nothing, say "No issues found" and list anything you were
unsure about. Do not present this as legal advice.

CONTENT:
[PASTE CONTENT]

The table output matters. When agents see the exact phrase next to a rewrite, they learn the pattern, and after a few weeks they catch it before the AI does.

A quick prompt for social captions

Social posts get written fast and posted faster, so here's a short version for your phone:

Fair housing check this caption. Flag any phrase that describes who
should live here, the neighbors, schools, safety, or anything tied to a
protected class. Give me a fixed version that only describes the
property. Caption: [PASTE]

What should you do when the AI flags something borderline?

When the AI marks something borderline, don't let it make the call, and don't make it alone either. Send it to your broker. Borderline items are exactly where state and local rules, and your brokerage's own policy, decide the answer.

Keep a running list of what your broker rules on. Add those phrases to the "flagged words" line in your prompt and your Project instructions. Over time your setup reflects your brokerage's real standards, not a generic list from the internet.

What to do this week

1. Pull up your last three listing descriptions and one recent listing ad or caption. Run each through the review prompt above.
2. Make a "Listings" Project in Claude and paste in the instructions, with your character limit and any local protected classes your broker gives you.
3. Ask your broker for the brokerage's flagged-word list and their policy on schools and distance phrases. Add both to the Project.
4. Check Meta's Business Help Center for the current housing ad category rules before your next listing ad.

If you want the listing tools I use in class, including the Listing Description Machine, they're in the free Claude 101 kit. For more ready-to-use prompts, see Claude prompts for real estate agents.