Ed 
Your listing description is the dating profile of your house. AI can write a passable one in 30 seconds. The problem is “passable” isn’t the same as “sells.” And there is a fair housing minefield that ChatGPT will walk you right into without ever telling you it happened.
I’ve been writing listing copy for almost 20 years, and I started running every draft through ChatGPT about a year ago to see what it gives me. Some of it is genuinely useful. Some of it would’ve gotten me a complaint filed with HUD before the listing was even live. Lets break down what works, what doesn’t, and the prompts I actually use.
What AI Does Well
The first draft. Honestly that’s the big one. Staring at a blank page is the hardest part of writing anything, and AI eliminates it. You feed it the basics (4 bed, 3 bath, 2,400 sqft, updated kitchen, pool, cul-de-sac lot) and 30 seconds later you have something readable to work with. Is it the final version? No. But editing is a lot easier than writing from scratch.
It’s also good at translating dry MLS specs into prose. “3 BR, 2 BA, 1850 sqft, 2 car garage, built 2008, .25 acre lot” reads like a Craigslist post. AI turns it into actual sentences a human would read. Not great sentences. But sentences.
It’s useful for giving you three tone variations so you can pick. Ask for one professional, one warm, one punchy, and you’ll usually find the right register hiding in one of them. I do this all the time when a property has an unusual vibe and I can’t quite land on the voice.
And it catches weak headlines. If your draft headline is “Beautiful home in great location,” AI will write you something better just by being asked. The bar is low (those headlines are everywhere) but the improvement is real.
Where AI Will Quietly Get You Sued
Fair housing law has seven protected classes: race, color, religion, sex, disability, familial status, and national origin. Listing copy can’t signal a preference for OR against any of them. The legal standard isn’t “did you discriminate” right, it’s “did the copy indicate a preference.” That’s a much lower bar. (Source: 42 U.S.C. § 3604(c).)
AI doesn’t know this rule. AI was trained on millions of listings written before the rules tightened, and it will happily reproduce the same violations everyone else made. Here are the phrases ChatGPT generates constantly that will land you in trouble:
“Great for families.” Familial status is protected. You can’t prefer buyers with kids, you can’t prefer buyers without them. Same with “perfect for empty nesters” or “ideal for newlyweds.” Any phrase that signals who SHOULD live there based on household composition is a problem.
“Walking distance to St. Mary’s Catholic Church.” Religion is protected. Mentioning a specific religious institution as a feature is steering. You can say “walkable neighborhood” or “close to local amenities.” You can’t name the church, synagogue, mosque, or temple as a selling point.
“Quiet neighborhood, no kids running around.” Same problem as “great for families,” just inverted. The fact that you’re advertising the ABSENCE of children doesn’t make it okay. It makes it worse.
“Master bedroom.” This one is less about HUD and more about MLS rules, but many MLSs (including ours here in Austin) now flag this term and ask you to use “primary bedroom” instead. The history of the word “master” in property law is the reason. Most MLSs will let it through with a warning, some won’t. AI defaults to “master” because that’s what 95% of its training data used.
“Safe neighborhood.” This is the trickiest one. In April 2026, HUD clarified that sharing factual crime statistics is NOT a fair housing violation. But the WORD “safe” gets flagged by MLSs anyway because it can function as a proxy for race or socioeconomic class. If you want to discuss safety, cite actual data (crime rates per 1,000 residents, for example) and not a feel-good adjective.
“Walkable for seniors.” Age can intersect with disability and familial status, and any language that suggests the home is suitable for ONE age group implicitly excludes others. “Single-level living” describes the house. “Perfect for seniors” describes the buyer.
AI doesn’t know it’s doing any of this. It just generates the kind of warm, friendly copy it’s seen a million times. The compliance layer has to come from a human who knows the rules.
Other Ways AI Will Embarrass You
Fair housing isn’t the only landmine. AI also hallucinates. I’ve had ChatGPT add a “wine cellar” to a property that didn’t have one, invent a “chef’s kitchen with Viking appliances” when the kitchen had a GE range, and describe a “spacious backyard with mature oaks” on a lot that backed up to a parking lot. It’s not lying on purpose. It’s just pattern-matching to what listing descriptions usually say.
It also reaches for the same tired phrases everyone else uses. “Won’t last long.” “Must see to appreciate.” “A rare opportunity.” These are filler. Buyers ignore them. Worse, they signal that the listing is boilerplate, which makes the buyer assume the house is too.
If you ask AI to look up your address and pull specs, it will sometimes invent the year built, the square footage, or the lot size. Don’t let it pull facts. Feed it facts.
And watch for made-up neighborhood names. Austin has a lot of unofficial sub-neighborhoods, and AI will sometimes generate plausible-sounding but fictional names. If you can’t find it on a county map, don’t put it in the listing.
One more: certain pricing phrases violate MLS rules. “Priced to sell” can be flagged as misleading in some MLSs. “Below market value” is even worse, because that’s a claim you have to be able to back up. AI uses these phrases reflexively. Strip them.
5 Prompts That Work
Here are the prompts I actually use, copy and paste. They work because they constrain AI in the ways it needs to be constrained.
1. The first draft prompt. “Write a 200-word listing description for a home with these features: [list]. Voice: warm but professional. Do NOT use any language related to family status, religion, age, ability, race, or safety. Do NOT use the phrases ‘won’t last long,’ ‘must see,’ ‘rare opportunity,’ or ‘priced to sell.'”
2. The headline generator. “Give me 5 headline options for a Texas Hill Country home with [features]. Each under 70 characters. No exclamation points. No phrases about who the home is ‘perfect for.'”
3. The translator. “Translate these MLS bullet points into 150 words of prose. No fluff, no filler, no superlatives. [paste bullets]”
4. The compliance rewrite. “Rewrite this listing description to remove any language that could be a fair housing violation. Flag what you changed and why. [paste description]” This is great even on copy YOU wrote, because we all have blind spots.
5. The gap finder. “What is missing from this listing description that a buyer would want to know? [paste]” AI is actually pretty good at this because it has read so many listings, it knows what readers expect.
3 Prompts That Will Get You in Trouble
“Write a listing description that appeals to young families.” Familial status. Do not do it. Even if your AI politely declines (some do, some don’t), the version that slips through will read like a fair housing complaint waiting to happen.
“Make this sound like it’s in a safe neighborhood.” Proxy language. AI will generate phrases like “established neighborhood,” “well-maintained area,” “peaceful streets,” which sound innocent but are widely understood as code. If the data supports a safety claim, cite the data. Don’t ask AI to imply it.
“Look up this address and write the listing.” Hallucination central. AI will invent features, get the year built wrong, and miss the actual selling points. Always feed it your specs. Never let it source them.
My Workflow as a Realtor
Here’s how I actually use AI when I write a listing. It isn’t “AI writes, I publish.” It is more like AI is the junior associate handing me a rough draft.
I start by giving AI the bullet points (specs, upgrades, neighborhood facts, unique selling points). I ask for three drafts in three different tones. I pick the one closest to right and ignore the other two.
Then I edit. I read every sentence and ask three questions: is it true, is it compliant, and does it sound like a human wrote it. About half the sentences get rewritten. Some of them get cut entirely.
Then I read it out loud. This is the step that catches AI-speak immediately. If a sentence has “nestled in” or “boasts” or “showcases,” it goes. AI loves those words. Real people don’t say them.
Then I check the headline and the first sentence against the fair housing list. These are the high-visibility spots where a violation does the most damage.
And I never publish unedited. Not once. The reason is simple, when something goes wrong with a listing description, it’s MY name on it, not OpenAI’s. I’m the one with the license. The buck stops here.
Why Pricing Headlines Are Different
One thing AI can’t do for you is price your home. The listing description is content, the price is a number tied to your local market. AI doesn’t know what comparable homes in your zip code closed for last month. It doesn’t know which ones sat for 90 days and dropped twice. It doesn’t know that the house two blocks over with the same floor plan went 40K over ask because three buyers showed up at the open house.
If you want AI to help with pricing strategy, it needs real comp data. That is why we built the Austin MLS MCP, which lets AI tools pull live MLS data when you ask pricing questions. (Same data on our listing pages, just delivered programmatically.) Without that, ChatGPT is guessing. We covered this in more depth in our test of ChatGPT as a home valuation tool, and the results weren’t great.
The Bottom Line
AI is a tool. It writes a fine first draft. It saves you time. It won’t catch its own fair housing violations, it will hallucinate features your house doesn’t have, and it will quietly reach for the same tired phrases every other listing uses.
Use it for the draft. Use a human for the polish, the compliance, and the pricing strategy. That’s the whole answer.
If you want to go deeper on AI in real estate, our pillars cover the full process: how to buy a house with AI and how to sell a house with AI. We also wrote about whether AI can read buyer offers, which is the flip side of this same question.
Frequently Asked Questions
Want a Realtor Who Writes Your Listing With Real Tools?
I use AI every day, but I also know where it breaks. If you’re thinking about selling and you want a listing that converts buyers without exposing you to fair housing risk, reach out. We use the same AI tools you would, plus 19 years of writing copy that actually moves houses in this market.
Neuhaus Realty Group covers Austin, Westlake, Lakeway, Bee Cave, Dripping Springs, and the rest of the Hill Country.
