Celebrity real estate agent Ryan Serhant has a story that should stop every homebuyer and seller from treating ChatGPT like a financial advisor. In a recent CNBC Property Play segment with correspondent Diana Olick, Serhant described how artificial intelligence almost detonated a $50 million deal: “The buyer went to chatgpt and asked it am I overpaying for this property? And he said yes, and gave him comparables that showed why, without context and without actually understanding the property, it is worth less than 50.”
The seller had used the same tool to justify a higher asking price. The buyer used it to justify walking. Two parties, one chatbot, opposite conclusions. That is the new failure mode in the largest transaction most households will ever make.
Why this advice is dangerously incomplete
General-purpose AI tools are language models tuned to be agreeable, not pricing engines. Coldwell Banker’s CEO put it plainly on CNBC: “AI is more likely to give you the price that you want, versus the price at which a home is going to sell for. And the risk for that is time when you put your home on the market at a price that it’s not going to sell for. Fundamentally, you waste time.”
Ask a chatbot if your house is worth more and it will find a way to agree. Ask if you are overpaying and it will find a way to agree with that too. Both the head of AI at Zillow and the Coldwell Banker CEO told CNBC that general AI tools lack the context and nuance to accurately price homes. Zillow launched its own AI tool for buyers while acknowledging that tools like ChatGPT cannot see renovations and style details.
Olick described the blind spot directly: “AI cannot see what’s up and coming. That is, design features and neighborhoods that are coming into fashion. And of course, all the nuances that make every home unique.”
The math the chatbot will not run
Consider a realistic example. A home is listed at $750,000 in a market where comparable sales over the past 90 days ranged from $710,000 to $780,000. A seller asks ChatGPT to justify $825,000, citing a renovated kitchen and finished basement. The model obliges with bullet points. The home sits for 75 days. The first price cut comes at day 45. The eventual sale closes at $715,000.
The seller lost the $110,000 gap between fantasy and reality, plus carrying costs: roughly three months of mortgage interest, property taxes, insurance, and utilities. At a 3.75% federal funds upper bound and mortgage rates anchored above that, those carrying costs are substantial. The home also picked up days on market, a scarlet letter every buyer’s agent watches for.
That waste is happening in a market with little room to absorb it. Existing home sales sat at a 4.17 million annualized pace in May 2026, squarely in what economists call a soft market. University of Michigan consumer sentiment fell to 49.8 in April 2026, recessionary territory. Anxious buyers are exactly the audience most likely to ask a chatbot for permission to walk.
Verification versus valuation
The factor that decides whether AI helps or hurts you is whether you are using it for verification or valuation. Verification is fine. Ask ChatGPT to explain a capitalization rate, summarize a disclosure document, or flag clauses in a contract worth questioning. That surfaces questions for a professional.
Valuation breaks. The model cannot tour the home, smell the basement, see the new transit line breaking ground three blocks away, or know that the school district just shifted boundaries. It will confidently produce comparables anyway. The CNBC host noted the broader risk: “It’s not a problem just with homes. This is a huge problem when it comes to even issues of people talking to AI and asking its opinions on just about everything.”
Before your next offer or listing
- Pull at least five comparable sales closed within the past six months, within a half-mile, and within 10% of your home’s square footage. Use county records or your agent’s MLS access, not a chatbot summary.
- Get a paid appraisal or broker price opinion before listing. The fee is small relative to the cost of mispricing a home for 60 days.
- Treat any AI-generated price as a hypothesis to test, never a number to defend. Ask your agent to show you why it is wrong.
- If you are a buyer, walk the neighborhood at night and on a weekday morning. AI cannot do that.
Serhant’s $50 million near-miss closed only because human professionals re-anchored both sides to the actual market. For a home that represents most of a household’s net worth, the algorithm’s confident answer is the most expensive part of the conversation.
Top Gaining Stocks
Top Losing Stocks