And why the gap between what the tool says and what the market pays is widening in both worlds.
A new survey from HomeLight landed in my inbox this week via the Kansas City Business Journal, and while the headline is about residential real estate, every number in it maps directly to what I see on the commercial side.
The finding: 64% of real estate agents say sellers are most likely to misunderstand what their property is worth based on information they got from AI and online tools.
Eighty percent of agents say accurate pricing from the start is the single most important factor in a sale. And yet, 50% of agents say home value is the expectation they most often have to reset with sellers.
The AI is not lying to sellers. It is doing something more dangerous. It is agreeing with them.
Todd Rhodes, HomeLight's VP of sales, put it plainly: AI models are designed to be helpful and agreeable, and they often lack the pushback required to set the right price.
That is a precise description of the problem. When a seller asks an AI what their house is worth, the AI reaches for averages, trending data, and comparable sales. It cannot feel the pressure a buyer feels when standing in the kitchen noticing the 2009 countertops. It cannot read the hesitation at the offer table. And critically, it cannot replicate the conversation where a good agent tells a seller something they do not want to hear.
Forty-seven percent of agents say when buyers are touring but not making offers, the asking price is most likely too high. Fifty-eight percent say a price reduction is the first recommended move when inquiries go quiet. Eighty-three percent say if a home is not selling, something has to change within four weeks.
Those are not algorithm outputs. That is market feedback. And the AI never gets that feedback loop because it is not in the transaction.
I see the commercial equivalent of this every week.
A property owner calls with a broker opinion of value they either generated themselves using an online cap rate calculator, pulled from an AI tool, or received from someone who had no business issuing it. The number is almost always the same: whatever the owner needs it to be.
Here is what those informal BOVs typically get wrong.
They use the wrong cap rate. An owner who bought in 2019 at a 6.5% cap rate does not want to hear that the market has repriced to 7.5% or 8% on their asset class. The AI does not push back on this. It finds a comp, applies the cap rate the owner fed it, and confirms the number.
They ignore lease structure. A single-net lease versus a triple-net lease on an identical building in the same submarket can move value by 15 to 20%. A modified gross lease with a landlord who has been quietly absorbing operating expenses is not a NNN deal, regardless of what the cap rate math says.
They anchor to 2021 and 2022 comps. The residential article makes this point explicitly: sellers are still mentally living in the red-hot market. Commercial owners do the same thing. That industrial deal that traded at a 5.2% cap in Q2 2022 is not a relevant comparable in October 2026 with the 10-year Treasury at 5.1%.
They ignore time on the basis calculation. In residential, 47% of agents say buyers who tour but do not offer are telling you the price is wrong. In commercial, the equivalent signal is NDA count without LOI conversion. If 30 qualified buyers signed a confidentiality agreement and none submitted an offer, the offering price is the answer to the question you are not asking.
They do not account for the capital markets environment. The residential article cites mortgage rates hitting a one-year high of 6.54%. In commercial, that pressure shows up as widening cap rate spreads, tighter lender underwriting, and buyers who cannot pencil the deal at the seller's basis. The AI that produced the owner's BOV did not account for what a regional bank's current DSCR requirements look like on a modified gross office building with 25% vacancy.
Residential seller concessions hit 44.7% of U.S. home sales in August, the highest rate since 2020. Median sale price in July was $407,730, which sounds stable, but that number is held up by the upper end of the market. The middle is being squeezed.
On the commercial side, U.S. CRE transaction volume in H1 2026 was $122.4 billion, up 11.77% from H1 2025. Industrial is leading. Distress is contained but not absent. Office CMBS delinquency is at 13.2%, the highest this decade. The 10-year hit 5.1% in late September after the Fed hiked again to 3.75% to 4.00%.
That is the environment a seller is pricing into. Not the environment their AI tool was trained on.
This applies whether you own a $450,000 single-family rental in Lee's Summit or a $3.5 million industrial site in the I-49 corridor.
Get a market-calibrated opinion, not a data-aggregated one. A real BOV accounts for current buyer pool depth, actual lender underwriting conditions, comparable closings from the last 90 days, and a frank conversation about what the market will bear in the current rate environment. It is not a cap rate calculator with a property address plugged in.
Separate what you need from what the market will pay. These are two different numbers. Conflating them is how properties sit for 120 days, price reduce twice, and close at less than an accurate original price would have generated. I have seen this cycle play out on both residential and commercial assets. The AI did not cause the 120-day sit. The AI-anchored expectation did.
Listen to the market signal, not the tool. In residential, the signal is buyers touring without offering. In commercial, it is NDA-to-LOI conversion rates, call volume drop-off after the OM goes out, and the first 30 days of market feedback. When the signal conflicts with the AI output, the signal is right.
Price for competition, not for aspiration. At 2001 E. 19th Street in Lawrence, we ran a 14-day call-for-offers and generated 7 fully underwritten competing offers. The property closed above asking. That result came from accurate pricing that created urgency, not aspirational pricing that created silence. The difference is not luck. It is whether you trust the market or the tool.
AI is a research tool. It is not a pricing authority. It can surface data, identify patterns, and summarize market trends faster than any analyst. What it cannot do is tell a seller the truth about their own asset in a way that lands, holds, and changes behavior.
That still requires a person who has been in enough transactions to know what the data is not saying, and who is willing to have the uncomfortable conversation before the market has it for you.
The 44.7% concession rate in residential is the market having that conversation. The 120-day sit with two price reductions is the market having that conversation. On the commercial side, the bid-ask spread that killed your deal in due diligence is the market having that conversation.
Better to have it with your broker at the listing table than with an expired listing in your rearview mirror.
Logan Freeman is the Managing Broker of Midwest CRE Advisors, based in Overland Park, Kansas. He has closed more than $450 million in commercial real estate transactions across the Midwest. Subscribe to Beyond Buildings: The F Factor for weekly market intelligence, deal breakdowns, and straight talk on the commercial real estate market.
Copyright 2026, Midwest CRE Advisors
