Industry
AI in Real Estate Advertising: Powerful Tool, Real Liability
Published August 24, 2026
Quick Answer
My job is to present a listing in its best light. That is what a seller hires me to do. AI has made parts of that job faster and better, particularly copywriting, image work and market research.
The problem is that the distance between best light and misrepresentation has gotten very short. Unreviewed AI output can produce fair housing language, inaccurate market figures, and photographs that show a property that does not exist.
When that happens, the exposure does not sit with the software. It sits with the listing agent and the seller.
I have been building and using technology in this business since well before it was fashionable. I spent years on the product side of real estate software before coming back to practice full time. So I am not here to tell you AI is dangerous and you should avoid it.
I use it every day. It is genuinely good at several things that used to eat hours.
But I also chair an ethics committee, and I read listings for a living. What I am seeing in the market right now concerns me, because the failures are not obvious. They look like good marketing. That is exactly what makes them a problem.
Watch: Mike Simoni, Bobby McShane and I talk through where AI helps in real estate and where it starts creating exposure for buyers, sellers and agents.
Market and pricing analysis
This is where AI has been most useful to me, and where it has also been most confidently wrong.
What works. Feed a language model a clean MLS export and it will sort, group and summarize faster than any human. It catches patterns across hundreds of records that I would need an afternoon to find. It is very good at organizing information.
What does not. It does not understand real estate math the way a practitioner does, and it will not tell you when it has guessed at a definition.
Here is a real example from my own work. I was calculating absorption rate, which is the measure of how quickly available inventory is selling. The model divided ninety days of closings by current inventory and reported that as the monthly absorption rate.
That is wrong, and the reason is a unit mismatch. Ninety days of closings is three months of sales. To get a monthly pace you have to divide that total by three before you divide by inventory.
What the error looks like in plain numbers
Say a town has 100 homes for sale and 60 closings over the last ninety days.
Wrong: 60 divided by 100 gives a 60% absorption rate, which reads as a strong seller's market with under two months of supply.
Right: 60 closings over three months is a pace of 20 per month. 20 divided by 100 gives a 20% absorption rate, which is a balanced market leaning toward buyers, with five months of supply.
Same data. Three times off. And it errs in the direction that tells a seller to price aggressively into a market that will not support it.
Think of it like measuring your speed by driving 180 miles in three hours and calling it 180 miles per hour. The arithmetic is fine. The unit is wrong, and nothing downstream survives it.
There is a second timing issue worth knowing. MLS PIN settled dates lag off market dates by roughly three weeks, so a trailing ninety day window that ends today is always missing sales that have closed but have not posted yet. I clamp every sales figure to a confirmed calendar quarter for exactly that reason.
None of this is a reason to stop using these tools for analysis. It is a reason to know your own formulas well enough to catch it. I have corrected models on this repeatedly. If I did not already know the right answer, I would have published the wrong one.
Listing copy
Agents are not professional copywriters. I will say that as someone with a marketing degree who still finds blank pages difficult. Having something to react to instead of nothing is a real advantage, and AI is a legitimately useful first draft machine.
Then there is the part that should worry every listing agent in the country.
Give a language model free rein on a listing description and it will write about raising a family in the home. It will mention the walk to a specific house of worship. It will describe the ideal buyer. It does this because it learned from decades of real estate copy that did exactly that, and much of that copy was written before anyone was paying attention.
Fair housing law does not care that a machine wrote it. It cares that it was published, and it was published under your license and on your seller's listing.
The failure mode is subtle, which is the danger. Nobody types a slur. What happens is that the copy starts describing who belongs in the home rather than what the home is. Familial status is the one I see most, and it almost always arrives wrapped in warm, pleasant language that sounds like good marketing.
The fix is not complicated. Describe the property, not the buyer. Rooms, systems, lot, light, layout, location. If a sentence could be rewritten to start with "perfect for," look hard at it.
Email and client communication
What works. Drafting, organizing, summarizing a long thread, catching what I forgot to answer. Nobody is harmed by a faster first draft.
What does not. Sending it as written.
People can tell. In 2026 the tells are common knowledge, and the average consumer now recognizes them faster than the average agent does. The rhythm goes flat. Everything arrives in threes. The em dashes pile up. Sentences restate the sentence before them in slightly different words.
I have a hard rule against em dashes in anything published under my name, and that rule exists for exactly this reason.
There is a more serious version of this problem than tone. If you have an assistant tool responding on your behalf, you are now sending messages you have not read. In a negotiation, that is not a style issue. That is your client's position being communicated by something that does not know what your client actually wants.
Negotiating with a computer does not build confidence. It does the opposite, and buyers' agents notice.
Photography and virtual staging
This is the one that generates the most argument, and it is where I see the clearest violations.
Start with why anyone stages at all, because the underlying premise is sound. According to the National Association of REALTORS® 2025 Profile of Home Staging, 83% of buyers' agents said staging made it easier for a buyer to envision a property as their future home, 58% said it positively influenced most buyers' view of a home, and 49% of sellers' agents saw staging reduce time on market.
You will also see much larger figures circulating, generally sourced to the Real Estate Staging Association. Those numbers are worth reading with the knowledge that RESA is the staging industry's own trade association and its comparisons are not controlled for price, cleaning or photography quality. The NAR survey asks REALTORS®, which makes it the more conservative and more defensible citation. Staging works. It just does not work as dramatically as the marketing around staging suggests.
Physical staging is expensive and slow. Virtual staging used to mean sending photos to a company where a person spent hours in Photoshop. Now every agent with a subscription can furnish an empty room in seconds, show it in four design styles, and generate before and after pairs.
That is a genuine advance. It is also where the rules start getting broken.
What MLS PIN actually requires
The rule is not ambiguous. An image is acceptable for filing with the service only if it contains, displays and is limited to objective information about the physical attributes of the property itself or about its location.
A furnished photo of an empty room, presented without a label, is not objective information about the physical attributes of that property. The room does not look like that. Disclosure is what makes virtual staging legitimate, and leaving it off is what turns a marketing tool into a rule violation.
I recently came across a listing that was clearly virtually staged with no disclosure anywhere on the images or in the remarks. That alone is a problem. What made it worse was what else had changed between the before and after photos.
Where it crosses into Article 2
Adding a sofa to an empty living room and adding quartz countertops to a kitchen are not the same act.
Furniture leaves with the seller. Countertops, cabinets, appliances and fixtures convey with the property. When an image alters those, it is no longer helping a buyer imagine their own belongings in a space. It is showing them a kitchen they are not buying.
Article 2 of the Code of Ethics requires REALTORS® to avoid exaggeration, misrepresentation, or concealment of pertinent facts relating to the property. Article 12 requires that we present a true picture in our advertising. An enhanced image that swaps out what conveys sits squarely against both.
Mike Simoni brought up an example in the video that I keep thinking about, because it captures how careless this can get. A patio had been generated into an exterior photo, and it was rendered sitting on top of the staircase leading to the back door. Nobody caught it. It went live.
That one is almost funny until you consider that a buyer somewhere scheduled a showing based on it.
My own buyers have walked into homes that did not resemble the photos. Nobody is happy in that moment. The showing is wasted, the seller loses a real prospect, and the listing agent has spent credibility they will need later.
The two tables
Here is how I sort it in practice.
| Use | What it does well | The condition |
|---|---|---|
| Market data analysis | Sorts, groups and summarizes large MLS exports quickly. Surfaces patterns across hundreds of records. | You define every time period explicitly and check the math yourself. |
| Marketing copy | Produces a usable first draft so nobody starts at a blank page. Good at varying structure and tone. | Describe the property, never the buyer. A human reads every line before it goes live. |
| Image enhancement | Corrects exposure, white balance and lens distortion. Makes a dim room look like it does in person. | Correction only. Nothing added, nothing removed, nothing replaced. |
| Virtual staging | Furnishes vacant rooms at a fraction of the cost of physical staging. Shows several design directions. | Labeled on the image itself. Furniture and decor only. Nothing that conveys gets altered. |
| Scheduling | Coordinates showings, sends reminders, keeps a transaction calendar accurate. | Lowest risk use on this list. Automate freely. |
Every row above is safe only because of the condition in the third column.
| Exposure | How it happens | Who it lands on |
|---|---|---|
| Incorrect data | Confident output built on an undefined time period or a misapplied formula. Absorption rate and days on market are the usual suspects. | The agent who published it, and the seller who priced off it. |
| Fair housing violations | Copy that describes who should live in the home. Familial status language is the most common, and it reads as warm rather than exclusionary. | The agent, the brokerage and the seller. |
| Misrepresentation | Undisclosed virtual staging, altered fixtures, generated features. Article 2 and Article 12, plus MLS PIN image rules. | The listing agent, with the seller exposed on disclosure. |
| Invented facts | Square footage, school information, permit history or tax figures that the model produced rather than retrieved. | Whoever put their license number on the advertisement. |
| Messages you never read | Automated responses negotiating or committing on your behalf in a thread you have not opened. | The agent, and any client whose position was communicated for them. |
Notice that the seller appears in four of five rows. This is not only an agent problem.
How we handle it
Our rules are short.
The standard we hold ourselves to
- A person reads everything. Every listing description, every email, every figure. No exceptions and no volume excuse.
- Photo edits correct, they do not create. Light, color and lens. If the camera did not see it, it does not go in the photo.
- Virtual staging gets labeled on the image. Not buried in remarks where nobody reads it.
- Nothing that conveys ever changes. Countertops, cabinets, appliances, flooring, fixtures. Not once.
- Numbers get verified at the source. If I cannot trace a figure back to an MLS PIN export, it does not go in front of a client.
- Nothing negotiates on my behalf. Drafting is fine. Sending is mine.
What sellers should be asking
If you are interviewing listing agents, these questions are worth asking, and the answers tell you a great deal.
Ask whether they use AI in their marketing, and watch whether the honest answer comes easily. Almost everyone does now. The ones who say no are either behind or not being straight with you.
Then ask what gets reviewed before it publishes, whether virtual staging is disclosed on the image, and whether anything that conveys with the house ever gets altered in a photo. Ask who writes the market analysis and where the numbers come from.
You are not looking for someone who avoids these tools. You are looking for someone who can tell you exactly where they stop using them.
The bottom line
Presenting a home in its best light is the entire job. I want the photos to be beautiful. I want the copy to be compelling. That is what a seller pays for and I make no apology for it.
The line is that the home in the advertisement has to be the home the buyer walks into.
AI made it dramatically easier to cross that line without meaning to, and dramatically harder for a consumer to notice. That is a bad combination, and it is why this deserves more attention than it is getting.
Use the tools. Read the output. Put your name on it only when you would defend every piece of it in front of a hearing panel.
Thinking about selling?
If you want to see how we market a listing, including exactly where we use these tools and where we do not, let's talk. I am happy to walk you through it in detail.
Get In TouchSources
- National Association of REALTORS®, 2025 Profile of Home Staging, released May 2025, based on a February 2025 survey of 1,266 REALTORS®.
- Real Estate Staging Association (RESA), industry survey data. Noted in text as trade association research.
- National Association of REALTORS® Code of Ethics and Standards of Practice, Article 2 and Article 12.
- MLS Property Information Network, Rules and Regulations, Section 1.0(f), Filing of Images, and Attachment D, Image Submission Policy.
- Fair Housing Act, 42 U.S.C. 3604(c), prohibition on discriminatory advertising.
This article is general information and is not legal advice. It reflects my own observations and views and does not represent the position of any association, committee or brokerage. Examples described are illustrative and no property, party or licensee is identified. For questions about advertising compliance, fair housing obligations or MLS rules, consult your broker, your association and a Massachusetts real estate attorney.