🤖 Seller's AI Says $648K, Agent Says $620K — Now What?

UPGRADE WITH LIBERTAS & EXP REALTY

🎧 Listen on Apple Podcasts and Spotify!

Here's what happened to Kelly, one of the agents in coaching. She was mid-listing appointment when the seller slid a printout across the table. Their AI said the house was worth $648,000 — six comps pulled, net proceeds calculated down to the dollar.

Kelly walked them through her own numbers and recommended $620,000. The room went quiet. The husband tapped the printout. "This says 648. You're saying 620. Why should we believe you over this?"

That isn't really a question about price. It's a question about value — what do you know that the AI doesn't? Most agents get defensive right there and lose the listing. Kelly didn't. She walked out with a signed agreement.

Today we walk through exactly how she answered it, the specific things AI structurally can't see about a property, the one phrase you should never say to a seller holding an AI printout, and the prompt-sharing move that gets everyone starting from the same honest number instead of a biased one.

A lot of consumer confusion about home value predates AI. AI is just making it louder. Plenty of sellers — including sophisticated, high-net-worth ones — still reach for cost-per-square-foot as their mental shortcut for value. It's about the bluntest instrument available for assessing a property, and it's frequently the primary variable raw AI valuation tools lean on, because square footage and sale price are the two data points most readily available at scale.

Real value depends on far more than square footage: which direction the house faces, which side of the street it sits on, who the builder was (in communities with multiple builders, one builder's homes reliably sell for more, for reasons that don't always show up in public data), what's been remodeled and when, and hyper-local conditions like a utility easement running through a backyard or a railroad line at the edge of a neighborhood. None of that is in the comps AI is pulling.

Concrete illustration. A rental comp recently surfaced showing an 1,800-square-foot house — no basement, one-car garage — that closed at $540,000 in a neighborhood where houses of that size sold for roughly $100,000 a generation ago. The number looked wrong at first glance — until the local context explained it.

The house sits across from a park that had recently been fully rehabbed — new landscaping, updated amenities, the works. Older comps in the same data set reflected the park mid-construction, when it looked like a dirt lot. A person with local, current knowledge catches that immediately. AI pulling from a static dataset has no way to know the park exists in its finished state, let alone that it moved the neighborhood's perceived value. That's the kind of detail that separates a professional's number from an algorithm's number — and it's exactly what you point to when a seller challenges you with a printout.

Critical rule, even if you're right. Do not tell a seller their AI's number is wrong, outdated, or that "AI doesn't understand real estate." AI tools sound extremely confident — no hedging, no wishy-washy language — and that confidence is persuasive even when the underlying analysis is shallow.

Dismissing the tool outright reads as defensive, and defensive loses listings. Your value isn't your MLS login. It's that you've physically seen the comps, walked the competing listings, know what's actually been renovated versus staged to look renovated, and understand the buyer psychology in that specific micro-market. Say that instead of attacking the tool.

Here's the actual fix, and it's simple. AI tools tend to mirror the bias baked into how a question is asked. A seller who asks "my house has an extra bedroom and great finishes — is it worth more than the neighbors'?" gets an agreeable, flattering answer. A buyer who asks "am I overpaying?" gets a reassuring one either way. The tool isn't lying. It's answering the question it was actually asked.

The fix: ask your own AI to write an objective, unbiased valuation prompt — one that instructs the AI to use sold comps only, stay non-emotional, and account for the specific variables above — and share that exact prompt with your buyers and sellers before they write their own. When everyone runs the same neutral prompt, the numbers converge naturally.

And when a seller shows up with a number that doesn't match yours, the first question is simple: "Did you use the prompt I sent you?" If yes, you're both working from the same objective baseline and can talk through the real gap. If no, you now have a clean, non-confrontational way to show them what the same AI says when it isn't fed a leading question.

Consumer AI use in real estate isn't a future trend — it's already the baseline. Roughly 48% of buyers say they expect to use AI somewhere in their home-buying process this year. Sellers are doing the same — checking your listing agreement, your reviews, and their own price expectations against AI before you ever walk in the door. Assume this is happening even with centers of influence and past clients who've worked with you multiple times before.

And yet — despite all of it — human agents remain firmly the standard. 91% of sellers still hired an agent last year. FSBO transactions hit a record low of just 5%. A recent survey found 44% of buyers would pay extra specifically to have a human expert double-check what their AI told them. People want the tool. They also still want you — provided you show up prepared.

Sobering related stat: roughly 80% of sellers only talk to a single agent before listing. There typically isn't a second chance to correct a fumbled first impression. If you walk into that appointment unprepared for an AI printout, or defensive when challenged by one, you don't get a follow-up meeting to fix it.

The professional response is preparation, not reaction. Before every listing or buyer appointment, run your own AI research on the property and the immediate competition — recent sales, local nuances, anything that would explain a gap between a raw algorithmic number and reality — so you walk in already knowing exactly where and why your number differs from whatever printout might land on the table.

Related preview. A turnkey, pre-trained AI real estate team — CMA prep, listing brochures, even AI-assembled property visuals pulled from public imagery when no other photos exist — is in final testing and expected inside Premier Coaching soon. The goal is that most of the appointment-prep work described above happens automatically, well before you ever arrive.

None of it replaces you. It removes the busywork so you show up to every appointment already several steps ahead — which, per the stats above, is exactly what today's buyers and sellers now expect.

Five concrete moves:

One — build your objective valuation prompt today. Ask your AI to write a bias-free CMA prompt using sold comps only, non-emotional language, and local-nuance instructions. Save it as a template.

Two — start sharing that prompt proactively with every new buyer and seller before their first appointment, so you're both working from the same starting point.

Three — practice the "did you use the prompt I sent?" script for the moment a seller shows up with a conflicting AI number.

Four — retire any language that dismisses AI as wrong or unreliable. Replace it with specific, local, first-person knowledge you can point to instead.

Five — before every appointment, run your own quick AI research pass on the property and its immediate competition so you walk in already anticipating the gap.

The seller with the AI printout isn't your adversary — they're doing exactly what nearly half of all buyers and sellers now do before meeting an agent. The agents who get defensive about it lose the listing. The agents who validate the tool, explain what it structurally can't see, and get everyone working from the same unbiased prompt are the ones walking out with signed agreements — the way Kelly did.

Nobody is being replaced by AI in this business. The agents at risk are the ones who fumble the conversation, dismiss the tool their client trusts, and show up unprepared for a question they should have seen coming. Be ten steps ahead. Get to work.

Ready to stop guessing and start producing?

🎯 Start Premier Coaching (free trial): premiercoaching.com
💼 Build wealth with Tim's eXp team: whylibertas.com/harris
📲 Elite Coaching — text Tim directly: 512-758-0206

If you built the unbiased valuation prompt this week and shared it with every buyer and seller before their first appointment — how differently would your next "the AI says something different" conversation go?

— Tim & Julie Harris

📬 Thanks for reading Harris Real Estate Daily. Share this with a colleague who needs clarity about where the industry is headed.

Forwarded by a friend? Sign up with just one click here.

Share this issue.