Robert Levine had lived in his Cooper City, Florida home for 15 years. When he decided last year to sell it, the natural first step was to sit down with some real estate agents and get their read on the market.
He did that. He listened to what they had to say about pricing. And then he decided they were wrong.
None of the agents he met with gave him much confidence. They had estimated the home’s value at a figure Levine thought was too low, and when he pressed them on it, they could not give him a number they were prepared to stand behind.
So instead of hiring one of them, he opened ChatGPT on his phone, and he and his wife spent the drive home asking it everything.
By the time the conversation was over, he had a plan. Within five days of listing, he had a signed contract for $954,800, $100,000 more than what every real estate agent had told him the house was worth.
How Did This Start?
The idea did not begin as a serious experiment. Levine and his wife were on a long drive from south Florida to North Carolina over the holidays, and they used the time to ask ChatGPT questions about home selling, mostly to pass the time and think through the process.
“Are we capable of doing this?” they typed. “What’s the realistic timeline tactically?” The questions started broad and got more specific. By the time they arrived, the experiment had become a plan.
Levine is the CEO of ComOps, a strategic consulting firm that advises casinos and hospitality companies on how to use AI.
He is not a typical first-time ChatGPT user. But he is adamant that what he did does not require his level of technical familiarity.
“I’d recommend it to everyone,” he told Fortune. “ChatGPT is not coding. It is a conversation, and you’re going to have to have that conversation with a real estate professional if you want to go that direction anyway.”
What Did ChatGPT Actually Do?
The scope of what the AI handled was comprehensive. It started with pricing, researching comparable homes in the area and building a case for why listing at $100,000 above the agents’ estimates was justified.
“When we met with real estate agents they lacked confidence in pricing,” Levine said. “ChatGPT gave us more confidence in price points of where the market was going.”
From there it moved into preparation. ChatGPT walked Levine room by room through what to address before listing.
“We repainted a couple of rooms in the house because ChatGPT said that’s where you’re going to get the biggest return on investment,” he said.
It flagged curb appeal details that might seem minor but affect first impressions. “It pushed us through all of that, including small things that I would have never thought of,” Levine recalled. “The first impression is important. We hear that all the time about curb appeal. But also when they walk into the house, they don’t want to see scuffs on the wall.”
Then it built a full timeline, when to start packing, when to schedule showings, when to hold the open house.
It recommended listing on a Tuesday specifically, explaining the strategic reasoning behind that choice.
It wrote the listing description, designed the open house handout, and guided Levine through the process of getting the home onto the MLS, the Multiple Listing Service used by buyers’ agents nationwide.
It suggested when to schedule viewings around Levine’s busy work calendar. He ultimately showed the home to 15 prospective buyers. One third of them submitted offers.
And when it came time to put the deal in writing, ChatGPT drafted the initial contract.
The Numbers Behind The AI Sale
The listing went live on Tuesday, March 3. Within 72 hours, Levine had five offers on the table.
He held the open house that Saturday as planned. By Sunday morning, five days after listing, he had a signed contract at $954,800.
The home sold for one of the highest per-square-foot prices in the Cooper City market, according to Levine, despite not having the best view, the largest lot, or the most updated finishes in the area.
He estimates he saved approximately 3% of the total sale price by not paying a listing agent’s commission, on a sale close to $1 million, that is a meaningful sum.
He still hired a real estate attorney to review the final paperwork. That was the only human professional involved in the transaction.
“It exceeded our expectations,” Levine said.
What ChatGPT Couldn’t Do
Levine is careful not to overstate what happened. He acknowledges that every sale is different and that his went unusually smoothly.
The approach required him to be engaged at every step, he had to prompt the AI for each piece of guidance rather than hand off the process to an autonomous agent operating on its own.
The AI could not host the open house, could not shake a buyer’s hand, and obviously could not pack up 15 years of belongings.
It also could not anticipate what might go wrong in a more complicated sale, a difficult inspection, a buyer who pulls out at the last minute, a title dispute.
Those are the scenarios where an experienced agent’s judgment and relationships have historically justified the commission.
Levine says he did not set out to make a case against the real estate industry. “It doesn’t necessarily replace professionals,” he said. “But it does allow us all to have the ability to be more curious and to feel more confident in the decisions we’re making.”
What The Industry Is Saying
The story has generated significant reaction in real estate circles since NBC 6 South Florida first reported it on March 11, and it picked up national traction this week when Fortune covered it.
Industry analyst Mike Hahn, writing in the trade publication Inman, argued that the deeper disruption in the story is not about ChatGPT replacing agents, it is about what the combination of AI and a flat-fee MLS listing service reveals about the traditional real estate infrastructure.
“What the Levine property example tells us is that an MLS-only company + AI + real estate lawyer is all you need today to sell a house,” Hahn wrote.
His sharpest observation was about the MLS itself. “If the MLS becomes simply a gateway to syndication,” he wrote, “then quite frankly, its days are truly numbered.”
Online reaction has been divided. Some observers pointed out that five offers in the first 72 hours could mean the home was underpriced even at $954,800, not that ChatGPT is a superior pricing engine.
“If he had five offers on day one, he undervalued his home,” one critic wrote, arguing that a thorough competitive analysis by a professional might have pushed the number even higher.
Others were more blunt about what the story means for the industry, “Either you use ChatGPT directly, or you pay a realtor a percentage to use ChatGPT.”
Where Is Real Estate Headed?
Real estate agent commissions have been a source of tension in the industry for years. A landmark antitrust settlement in 2024 changed some of the rules around how buyer’s agent commissions are disclosed and negotiated, cracking open a market that had operated under a largely fixed-commission structure for decades.
That settlement created the conditions for exactly the kind of experiment Levine ran, a seller who felt equipped to bypass the listing agent entirely and keep that side of the commission for themselves.
Levine saved roughly 3% on a $954,800 sale. That is approximately $28,644 that stayed with his family rather than going to an agent who, in his estimation, was less confident about his home’s value than a free AI chatbot.
“The more you use AI,” he said, “the more confident you’ll become in leveraging those tools.”
