Teach it once instead of correcting it every time. Give the AI your three best listings as examples, write down your rules (facts only from your notes, nothing personal about the seller, describe the home and not who should live there), and save all of it as a reusable skill. Every new listing then comes back as MLS remarks, a buyer email and a fact check in your voice, and your job becomes a quick review (about ten minutes in my example) before your broker signs off.
Every agent I know writes listings a little differently, and every one of them thinks their way is the right way. Fair enough. Their way has won them business for years. AI doesn't know your way until you show it, so most agents end up showing it again on every single listing.
Most agents who use AI already use it for this. NAR's 2026 Technology Report puts listing descriptions as the top AI use, at 75% of the agents who use AI. The problem is doing it consistently. In my experience, most agents still go two or three rounds before they get something they'd put their name on.
Why does AI make every listing sound the same?
Because you're starting from zero every time. You know exactly how it goes. You paste in your notes and it comes back sounding like a hotel brochure. So you tell it to sound more like you. Make it shorter. Remind it the pool house out back only has a half bath. It gets better. Then the next listing comes in and you do the whole dance again.
None of those corrections stick. The fix is to write them down once, in a place the AI reads every time.
How do you build a listing skill?
A skill is a saved set of instructions plus examples that the AI pulls up whenever you ask for it. In the video I built mine in Claude. It's one prompt and about three minutes of the AI working.
| Step | What you do | Why it matters |
|---|---|---|
| 1 | Pick your three best listings | Good AI always needs a good example to run off of. It copies the quality it can see. |
| 2 | Paste them in with a short prompt asking it to build a listing skill | It studies how you write and turns that into instructions. |
| 3 | Read the voice section and fix anything that isn't you | This is the part that makes it sound like you and not like everyone else. |
| 4 | Add your rules and your broker's rules | Facts only from your notes, nothing about the seller, fair-housing wording, words you never use. |
| 5 | Save it, then test it on a listing you've already posted | Compare against the real one before you trust it on a new listing. |
Why three? It's my rule of thumb. Three great examples give it a clear picture of what good looks like. Go past three and, in my experience, the AI starts to get overwhelmed. If you only have two you love, use two.
What goes in the voice section?
Mine came back describing how I write as quiet and confident: the specifics do the selling so the adjectives don't have to. Then a few hard rules. My favorite is no em dashes. They're one of the biggest AI tells out there.
Then a list of words it never uses: stunning, breathtaking, nestled, luxurious, exquisite. You get the point. Every agent has their own list. Write yours down and the AI stops reaching for them.
What does the skill hand back on a new listing?
For the demo I ran it on a made-up listing: a fourteen-and-a-half-million-dollar canyon-view home where the seller wants total privacy. One sentence asking it to run my listing skill on the notes, and here's what came back.
| Output | What it's for |
|---|---|
| MLS remarks | The listing itself, written to match your three examples. Mine opened with the build year, the remodel and where the canyon views are. |
| Buyer email | A ready-to-edit email for your buyer list, in the same voice. |
| Fact check | Every claim lined up against your notes, so you can scan the list and see whether the facts match exactly. |
| What it left out | Things from your notes it chose not to use, and why. |
| Questions for the seller | Gaps it noticed. It asks instead of guessing. |
The draft itself is fast. In my test runs it came back in under a minute. The fact check and the "left out" list are where the skill earns its keep.
What should AI leave out of a listing?
Two things it caught in the demo. The notes said the seller was a film producer. That's personal and confidential, and most sellers don't want buyers knowing it, so it stayed out.
The notes also said the neighborhood was "mostly family, great for kids." It left that out too, and it should. That line describes who should live there instead of describing the home, and familial status (households with children) is a protected class under the federal Fair Housing Act. The California Department of Real Estate's March 2026 advisory on AI points to its regulation 2780, which bars ads that show a preference or limitation based on protected characteristics, and it says using AI doesn't shield a licensee from fair housing liability. Leave it out and keep it out.
Then there are the seller questions. Those are the fixes you'd normally make after the listing is already live. Now you get them before. Look through them, pick the ones that matter and reach out to the seller for more.
The same DRE advisory says licensees should avoid putting confidential or sensitive client information into public or unsecured AI platforms, and that brokers have to supervise the AI tools their agents use. Check your brokerage's AI policy first. Give the AI the property facts and keep the rest out.
Can you turn the listing into social posts too?
Yes, and this is where it gets fun. Once the listing skill works, you stack a second skill on top of it. I built a social skill the same way: my best posts go in as examples. Feed it the approved listing and it hands back a week of posts: a just listed post with the carousel, a short video to shoot and a check-before-posting list.
That last part matters. It flags anything you should double-check before it goes up, including making sure the confidential seller details are still out. One repeatable writing skill takes the back and forth out of AI. It becomes a recipe. Stacking a second one gets you a week of content from the same listing.
How much time does a listing skill actually save?
Quick math, with my example numbers. Say a listing takes you about an hour today. With the skill it's about ten minutes, and most of that is you reading it over. In my runs, the AI gets you roughly 85% of the way there and you button up the last 15%.
Your numbers will be different. Run them anyway. Listings per year times minutes saved is real time back in your work week.
Who checks the listing before it goes live?
You do, and your broker does. The review is the real ten minutes, and it should be. The AI doesn't know what it doesn't know, which is exactly why the skill asks instead of guessing. The DRE advisory says the same thing from the legal side: licensees using AI to draft listings have to independently verify the facts, and an AI mistake doesn't excuse the violation.
So the flow is simple. The skill writes the first draft and checks it against your notes. You read it. Your broker signs off. Then it goes live. Nothing posts without a person on it. Build it once, and every listing after that is a quick review.
This isn't legal advice. Follow your state's rules and your MLS's rules.
Frequently asked questions
Does this work in ChatGPT or only in Claude?
The video uses a Claude skill. The same instructions should work as a ChatGPT Project: paste them into the Project's instructions along with your three example listings. Test it once on a listing you've already posted and compare before you trust it on a new one.
Can I paste real seller details into the AI?
Check your brokerage's AI policy first. California's Department of Real Estate advises licensees not to put confidential or sensitive client information into public or unsecured AI platforms. Give the AI the property facts it needs and keep the rest out.
Will the AI keep my listings fair-housing compliant?
It helps, and it doesn't replace you or your broker. Put the rule in writing (describe the home, not who should live there) and the skill will flag lines like "great for kids" and leave them out. You still read every line, and your broker still signs off. This isn't legal advice; follow your state's rules and your MLS's rules.
How long does a listing take once the skill is built?
The draft came back in under a minute in my test runs. The real time is your review. My example: if a listing takes you an hour today, it comes down to about ten minutes, mostly reading. Plug in your own numbers.
Full video transcript
Every agent I know writes listings a little differently, and they all think their way is the right way. And fair enough. It's won them business and got them clients over the years. So let's teach AI your way to write things. After that, every listing goes from a bit of back and forth, which might take 30 minutes to an hour, down to about 10 minutes. I'm going to show you how you can cut that time right now.
If you're still going back and forth with ChatGPT on every listing and doing the same work twice, let me show you how to do it once, the right way. This is the AI setup I'd build for any agent tomorrow. It takes 10 minutes and it pays you back on every listing after.
The Realtors' tech report came out last month, and out of all the people using AI, three out of four are using it to write listings. So pretty much everybody's doing it. The problem is a lot of people aren't doing it consistently, and they're still doing it two or three times over before they get their quality outputs.
And you know exactly how that goes. You paste in your notes. It comes back sounding generic, like a hotel brochure. You tell it to sound more like you, maybe make it shorter, remind it that there's a pool house in the back that only has a half bath. And on the next listing, you do the same thing all over again.
So we're going to build this together, starting off with three of my best listings. I'm going to put them into this quick prompt here. It's going to help us develop a great skill. Obviously this is all demo data, but here's what good looks like to me. You all know what good looks like to you. So we have our three examples in here, and we're going to go ahead and run this.
Okay, this ran for about three minutes, and this is what returned. We have a listing skill here called "listing in my voice." It turns property notes into full listing packages that sound like I wrote them.
Looking down at this next section, this is how I write: quiet, confident, specifics do the selling so adjectives don't have to. One of the ones I love is no em dashes. Em dashes are big-time AI tells, so we try to keep those out. Things to never use: stunning, breathtaking, nestled, luxurious, exquisite. You get the point.
Here is what good looks like. Good AI always needs a good example to run off of, so having three great examples here is perfect. More than three and the AI starts to get overwhelmed.
All right, let's go give this skill a shot. We have a new listing. It's a fourteen-and-a-half-million-dollar canyon view, and the seller wants total privacy. We're going to say, "Please run my listing in my voice skill for this new listing."
Let's check this output. We have our MLS remarks first: "Built in 1962 and fully remodeled in 2019, this six-bedroom has canyon views from the primary suite and the back terrace." Sounds just like our good examples. Next up is our buyer email. This looks pretty solid.
Next up is our fact check list. This is a great way to look down the list and see whether the facts match exactly. Then what was left out, and this is really important. The seller was a film producer. Well, that's personal and confidential. Many sellers probably don't want potential buyers knowing these things. "Neighborhood's mostly family, great for kids." We might not want to say that.
And then we have some questions here for the seller that we can look through, see if any of them are pertinent, and reach out to the seller to get more info.
Next up, let's talk about stacking skills. We're going to take our listing skill and roll it into a social skill, so we make social media posts out of it. I've already created this skill, and here's what we get back: a week's worth of posts. We have a just listed post that gives us our carousel, and a short video we can go ahead and shoot. And there's a check before posting: the confidential seller details, making sure all of that is still correct.
So out of one repeatable writing skill, we're able to take out the back and forth with AI and put together something that's like a recipe. This is what good looks like. And then we're skill stacking on top of that and adding a week's worth of social posts.
So let's do some quick math. Say a listing takes you about an hour. With this, it's going to be about 10 minutes, and most of that is you reading it over. AI is going to get you about 85% of the way there, and you button up the last 15%. That's 50 minutes back on every listing. If you're doing 20 listings a year, that's over 16 hours a year you're getting back. Plug in your own numbers. That's real time back in your work week.
Now look, the draft takes 10 seconds. Your review is the real 10 minutes, and it should be. It doesn't know what it doesn't know. So instead of guessing, it asks me, and that's exactly what you want in this situation.
So the skill writes the first draft and checks it against your notes. You read it, your broker signs off, and then it goes live. Nothing posts without a person on it. Build it once, and every listing after that is just a quick review.
I'm going to give you both of these skills for free, the listing one and the social one. The links are down below. No email, no catch, just free value. This is a framework and a template. Every spot where your stuff goes is marked. Drop in your best three listings, your best three posts and your broker rules, and it's all yours. And if you'd rather have us build it with you, there's a link for that too.
Rather have it built with you?
We sit down with your best listings and your broker's rules and build the listing skill and the social skill together, set up the way your team actually works. If you'd rather do it yourself, both skill templates from the video are free.
Book a 30-minute fit call