I tried to sell my house with a chatbot
The single most important part of the process is setting the listing price, and agents say they play an essential role.
Even making AI more truthful or warning users about bias may not fully prevent harm, as chatbots can still subtly reinforce existing beliefs, the study finds. (Image for representation: FreePik) I was sitting in my car when I got a phone call that was potentially worth more than half a million dollars. It was a real estate agent whose client was considering making an offer on my house. She wanted to clarify a few details, including: Had I really handled the whole listing privately? She couldn’t quite believe that I was an amateur.
“So — you’re not a Realtor?” she asked.
“No,” I said. “It’s the first house I’ve ever sold.”
“I’ve been in this job for, well, more than a day, and I was sure you were a Realtor,” she said. “Everything — the language you used, the organization, the emails.”
But none of that was me. It was all AI.
A few days earlier, I had pressed go on an experiment involving my family’s single largest financial asset. Could I sell our home without a human real estate agent, relying almost entirely on a couple of chatbots?
As a technology journalist, I had watched artificial intelligence transform medicine, business and even warfare. But I didn’t know how it would function in the far more intricate world of Hudson Valley real estate.
“Can we really do that?” my wife asked. “We don’t know anything.”
“Just trust me” was the best I could come up with. “AI can handle it.”
Our three-bedroom, two-bath ranch was set on a marshy acre in upstate New York. My wife and I had bought the house for about $520,000 four years ago, and I quickly fell in love with its vaulted sunroom and giant kitchen island, where we’d gather for Cheerios every morning and the occasional Negroni at night. But in March, with a second child on the way, we decided it was time to move. In the beginning, we expected to do what 91% of people do when they sell a home: Hire a real estate agent.
The single most important part of the process is setting the listing price, and agents say they play an essential role — using their wealth of experience to synthesize market data with intangibles like local demand and neighborhood quality. My wife and I figured our house could fetch $550,000, but that was pure guesswork, based on nothing more than glancing at Zillow. We looked forward to the agents’ more sophisticated approach. Yet when we interviewed a few and mentioned that figure, they mostly just shrugged and agreed.
A few days later, I got curious about how our home would look to potential buyers, so I gave an AI chatbot the basic details and asked for a listing description. It generated glowing prose — describing an “inspiring” place that was “flooded with light” and “perfect for bird lovers.”
So much of a real estate agent’s job seemed straightforward: Follow the rules; take cues from comparable homes. Any human edge that agents may have once provided seemed archaic. They no longer have total control of listings or home sale data, and the personal networks that once helped a property sell have long since been obviated by websites like Zillow and Redfin.
With my wife’s support, I decided to replace the agent entirely with AI.
Once I started using AI, I couldn’t stop. To get our house ready for sale, I asked hundreds of questions over three weeks. When I needed to hire a photographer, the chatbot offered a list of local businesses, then gave advice on staging. It told me how to organize the resulting photo gallery for maximum impact. (Exterior > kitchen > floor plan.) As I obsessively fine-tuned the listing language, it indulged my every request.
Real estate agents used to tightly control access to the Multiple Listing Service, the master database of homes for sale. But a rule change more than 20 years ago allowed those listings to circulate far more easily online, and I used a service called Homecoin to publish mine for $200. The documents I needed to fill out were intimidating at first, but AI quickly explained jargon like “selling concessions” and “automated valuation model.”
The chatbot did make one major mistake. I wasn’t using an agent to sell, but I would still be expected to pay the eventual buyer’s agent a commission of up to 3%. When I complained about this to the chatbot, it said I should explain in the listing that I was offering 0%. That’s not allowed, under a 2024 settlement that would have exposed me to fines. Thankfully, I knew that already, because Homecoin had flagged it as a common mistake.
Aside from that snag, AI made the process of creating and submitting our listing feel easy. I arranged for it to be published Thursday, March 19. That morning, when I logged into Zillow, my home popped up right away. It blended in perfectly with the houses represented by real estate agents. It was impossible to tell that I had no idea what I was doing.
A flurry of bookings to view our house over the coming weekend arrived in my inbox within hours. I struggled to manage the appointments until, again, I just let the chatbot do everything for me. I told agents that they had to email or text — no phone calls. Whatever they wrote, I copied and pasted into the chatbot; whatever it replied, I copied and pasted right back to the agents. I was worried that pushy ones would prey on my inexperience, so I had the chatbot come up with a list of potential conflicts and write confident responses I could have ready.
It was now almost impossible for me to make a decision without getting AI’s opinion. By Friday evening, I was starting to worry that the interest in our house was a little too strong. We had nearly 20 viewings scheduled for the weekend. I confessed to the chatbot my anxiety that we had underpriced the home. It offered some needed reassurance, saying that by pricing low, I had stumbled into an “accidental strategy” that could result in multiple offers. “When you get 1,100 views and 91 saves, you haven’t just listed a house; you’ve started a localized ‘gold rush,’” it wrote.
On Monday, I took stock. After a weekend full of showings, we had lots of rejections. Everything from the neighbors to the price seemed like problems. One buyer said the house was too small; another complained it was too big.
The chatbot insisted that this did not matter. “I know hearing ‘no’ 14 times feels like a series of bad dates,” it wrote, “but in the real estate world, those numbers are a sign that your listing is healthy.” The chatbot was right. I got three offers by our 5 p.m. deadline. All were for a lot more than the asking price.
The chatbot raved about our shared victory. But we were entering the stage I dreaded the most: negotiations. I’m the kind of person who gets sweaty palms when asking for a better table at a restaurant, and I’m incapable of shushing someone in a movie theater. So how could I handle the high-stakes back-and-forth of a real estate sale, while my family’s life savings were on the line?
When I looked at our three offer letters, though, I saw something different. They were simple, one-page documents, and only a few things seemed to matter: mostly the price, conditions and deposit. All three bidders waived inspection and appraisal, and they all had healthy financing. I uploaded the offers to the chatbot. It spat out a verdict that my wife and I agreed with: The offer with the most certainty was the winner, even if the price was a smidgen lower than the highest.
Chatting with the bot about possible counteroffers, I struck upon an idea. Would the buyers agree to pay their own agent a 2% commission instead of making me pay? That would instantly save me more than $12,000. If I were using my own agent, making a commission the focus of a negotiation would feel awkward, even insulting. The commission is the lifeblood of the entire industry. But I had no agent to worry about, and my chatbot supported the idea. So I went for it. My counterparty accepted almost instantly.
We accepted the offer at just over $600,000. And my AI experiment came to a close. To handle the closing, I hired a lawyer — a human one — for a small fee.
In the end, using AI netted me more than $90,000. That includes the premium over the asking price, plus the roughly $36,000 in fees I didn’t pay.
I’m not sure how replicable my experience would be for anyone less versed than me in technology. I’m paid to be an expert on AI. Selling the house was a big and complex undertaking that worked in part because I was transacting in a thriving market, with no special circumstances. Many people would probably be happy to pay fees to avoid the hassle. Some will be too afraid to sell their home without a paid professional in their corner — someone who knows all the rules and, unlike a bot, is obligated to follow a code of conduct.
But I’m persuaded that AI may well transform real estate agents into something more like travel agents. Once essential to navigating an opaque process, they could soon become more of a nice-to-have for busy people who want a more carefree experience.
Selling my home for $605,000 without any agent fees put as much cash in my pocket as a $643,000 sale when paying agents their normal commission. Maybe I could have sold for more with a human’s help, but that wouldn’t have necessarily benefited me. (And it certainly wouldn’t have benefited the buyer.) The only people guaranteed to gain would have been the real estate agents.
Share
