ChatGPT nearly derailed a $50 million Manhattan penthouse sale
Ryan Serhant said a buyer tried to back out after ChatGPT questioned the price, forcing his brokerage to defend the deal with private market data.
By Daniel Okafor · Business Editor
3 min read
Ryan Serhant said a ChatGPT answer nearly broke up a $50 million New York City penthouse sale, showing how fast AI output can affect high-stakes negotiations. The celebrity broker and founder of Serhant told Fortune’s Brainstorm Tech conference that the deal survived only after his team used market information the chatbot could not see.
Serhant described the property as a trophy penthouse that was difficult to value because comparable sales were scarce, according to Fortune. The buyer and seller had already gone through a tense negotiation before the parties agreed to a deal sheet at an even $50 million, he said.
The problem came late in the process. Serhant said the buyer asked ChatGPT whether $50 million was too much to pay for the property, and the chatbot responded that it was.
According to Serhant, the buyer’s broker then called to say the buyer wanted out because AI had concluded the property was not worth the price. Serhant told Fortune’s Allie Garfinkle that he pushed back sharply, arguing that the buyer should rely on data rather than a generic chatbot answer.
The seller then used ChatGPT as well, Serhant said. His client asked the chatbot the opposite question: whether $50 million was too little after a buyer backed away because of ChatGPT’s advice. Serhant said ChatGPT agreed with that framing too.
Serhant said his firm rescued the transaction by returning to off-market research and context unavailable to large language models. He said those tools can reflect material found online, but they do not know private conversations, nonpublic market information or the direction of a deal.
He also posted a video about the dispute on social media, Fortune reported. Serhant said the video drew 3 million views in roughly three hours, both clients saw it, and the parties resumed talks before closing the deal.
Serhant used the episode to argue that AI has limits in luxury real estate. He said models can summarize the internet’s record but cannot account for information outside public sites such as Reddit, Zillow and Realtor.com.
The comments fit into a broader argument over whether AI will reduce the role of real estate agents or make them more useful. Fortune reported that Andrew C. Spieler, a distinguished professor in business and finance at Hofstra University, told the publication in March 2024 that agents increasingly resemble travel agents because consumers can now access information once controlled by professionals.
Spieler told Fortune that buyers once depended more heavily on agents because agents had access to listing data unavailable to consumers. With more information online, he argued, the role has changed.
Serhant took the other side at the conference. He said wealthy clients still want a person to guide decisions, take responsibility when problems arise and provide judgment that AI cannot supply. He summed up his view by saying people dislike being sold to but enjoy shopping with friends, according to Fortune.
This story draws on original reporting from Fortune.