AI in Real Estate: Why Your House Isn’t Selling

AI in Real Estate · Field Notes 01

There is a lot of noise around AI and real estate right now.

“Use it to write your listing description.” “Ask it for Instagram captions.” “Generate a month of content in thirty seconds.”

Fine. But if your house has been sitting on the market for six weeks without an offer, another clever caption is a long way from an answer.

I want to know why buyers are choosing something else.

That is where AI starts getting useful. Give it reliable information, make it challenge your assumptions, and have a real professional check what it finds. That process can help sellers, buyers, and agents make better decisions.

We’re using a Ferrari to get groceries

Writing is an easy place to start with AI. It can save time. But stopping there leaves a lot of the tool’s usefulness untouched.

Home search is already moving beyond a box full of filters. Zillow announced its conversational AI mode in March 2026, and Realtor.com announced RealAssist AI in June. Both announcements began with limited or beta access. They show where the tools are headed.

The useful question is what you can do with the information today.

Your house isn’t selling. Start with the evidence.

Here is a hypothetical example, not a report on an actual listing. A home has been available for six weeks. It has:

2,800 online views and 147 saves
11 showings, with no second showings
No offers
Two price reductions

A seller might say, “We need more marketing.” An agent might say, “We need another price reduction.” Either explanation could be right. Neither is proven by that list.

Views and saves tell us people encountered the listing. Showings tell us some people wanted a closer look. Those numbers alone do not reveal why anyone passed. The same person may view a listing repeatedly, platforms count activity differently, and eleven showings are still a small sample.

Use AI to organize possible explanations and identify what evidence would distinguish them.

Give it a dated, accurate information packet:

Asking price, listing history, and the dates and amounts of each price change
Verified property facts, including size, layout, condition, lot, taxes, and relevant fees
Current photographs, floor plan, and listing description
Showing history and actual buyer feedback, with client identities removed
Competing active homes, relevant pending listings, and recent comparable sales

Have your agent select the comparisons. In Northern Michigan, the kind of waterfront, access, condition, and location can change the comparison substantially. A similar bedroom count or a shared ZIP code is a starting point for investigation.

Then ask the uncomfortable question.

Make AI argue against your house

Asking AI to list your property’s strengths makes it easy to get a reassuring answer. A harder prompt can reveal assumptions you have stopped noticing:

“Act as a skeptical buyer with a $450,000 budget. Using only the dated listing information and competing properties I provide, build the strongest case for choosing another home over this one. Examine price, condition, location, photography, layout, and perceived value. Tie each concern to a specific fact or source. Separate verified facts from hypotheses. List missing information. Do not invent buyer feedback, comparable sales, or repair costs. Do not soften the answer to protect my feelings.”

Then follow up:

“Which three changes deserve investigation first? For each one, show the supporting evidence, the evidence that would challenge it, and the next step needed to verify it. Do not promise a sale or assume every problem requires a price cut.”

Maybe a competing home offers substantially better condition for a modest difference in price. Maybe the first photograph undersells the property. Maybe buyers cannot understand the floor plan. Maybe the description spends two hundred words on countertops and overlooks the feature that actually sets the home apart.

Or maybe the asking price simply asks too much of buyers when they compare their alternatives.

These are explanations to test. Review the original evidence, talk to the people who toured the home, and decide what deserves action. Buyer alternatives and market evidence should drive the pricing conversation.

A confident answer still needs checking

AI does not walk through the house. It does not smell the basement, hear the road noise, or know how the property feels on a wet February afternoon.

A general AI chat also does not automatically have access to your local MLS, complete disclosures, showing feedback, or the latest contract documents. Give it incomplete information and you can get a beautifully written explanation of the wrong problem.

Ask where each claim came from. Open the source. Check the date.

A model can agree too readily, miss context, or invent details. Asking it to be skeptical improves the question; it does not guarantee that the answer is correct. NAR’s discussion of agent trust in AI highlights the importance of oversight and accuracy.

Sellers and agents can become attached to a property. “We spent $8,000 on the flooring.” “Everybody loves the backyard.” “The house down the street sold for more.” Those statements deserve context. Buyers are comparing what they can get for their money today.

Make the comparison explicit. Then verify it.

Buyers can run the same exercise

If you are considering a $385,000 home, “Is this a good house?” is too vague to do much work.

Try this instead:

“I am considering this property at $385,000. Using the listing, disclosures, tax information, inspection findings, comparable sales, and alternatives I provide, build a case for buying it and a separate case for walking away. Identify the five most important unanswered questions. Cite the source for each factual claim, label assumptions, and tell me which professional should verify each issue.”

Take the questions to your agent, lender, inspector, insurance company, or attorney. A question about a possible structural problem needs an appropriately qualified professional. A financing question needs your lender’s assessment of the actual property and your situation.

AI can help you arrive at those conversations better prepared. For more on the questions to ask before committing, start with buying in Northern Michigan.

Agents: turn the saved time into better client work

There is a business opportunity here. It starts with doing useful work more consistently.

An agent can use AI to organize a client’s questions before a meeting, compare a supplied set of properties, summarize recurring showing feedback, or identify a promised follow-up that still needs attention. Each task should end with a checked answer or a concrete next action.

Here is a practical workflow:

Choose one problem. A stalled listing, an undecided buyer, or an unanswered client question.
Supply the evidence. Use current records you are permitted to use. Remove unnecessary personal information.
Ask for competing explanations. Require sources, missing facts, and questions that could prove an explanation wrong.
Review the result. Check it against the records, local knowledge, and the relevant professional’s findings.
Act and measure. Have the client conversation, agree on the next step, and record what actually happened.

If AI saves thirty minutes, use that time to answer someone properly, follow through on a promise, or prepare for an appointment. That is how it can contribute to a bigger business: better service, more useful conversations, and fewer loose ends.

Measure the time saved, the conversations that happened, the appointments created, and the decisions moved forward. Generating fifty messages proves you generated fifty messages. Revenue claims require actual results.

Keep the work honest

Use AI carefully with client information. Review the tool’s privacy settings and your brokerage’s policies, and use only information you have permission to process. A property comparison rarely needs someone’s tax return, bank statement, or personal identifier.

Keep property images truthful. Editing a photograph to hide damage or manufacture a view can mislead a buyer. Clearly identify conceptual artwork and any permitted virtual staging. NAR’s 2026 Code of Ethics, Article 12 and Standard of Practice 12-10, addresses accurate advertising and misleading content, including images.

The same Code’s Article 10 addresses discriminatory real-estate services and advertising. Keep recommendations focused on objective property features and the buyer’s stated needs. An algorithm does not remove an agent’s responsibility for what they recommend or publish.

Here is what I’m going to test

I’m starting a series on practical AI use in real estate: property comparisons, buyer questions, stalled listings, and the day-to-day work that helps an agent serve people better.

For each test, I want to show the question, the information provided, what the tool produced, what needed correction, and whether it helped anyone make progress. Examples will use hypothetical situations or information that can appropriately be shared.

Some tools will be useful. Some will waste time. I’ll show the difference.

Better information. Better questions. Better decisions.

Have a Northern Michigan real-estate question you want examined? Send me the question. We can start with the evidence and figure out what deserves a closer look.

Questions people ask about AI in real estate

Can AI tell me exactly why my home has not sold?

It can organize evidence and suggest explanations to investigate. It cannot reliably infer every buyer’s reason from views, saves, or showing counts. Your agent needs to check the actual feedback, comparable properties, presentation, and price position.

Does AI have access to the local MLS?

Access depends on the specific product and its authorized data connections. A general chat should not be assumed to have current or complete MLS information. Ask what data it used and when that data was updated.

Can AI and real-estate agents build a better business together?

AI can reduce preparation and organizational work so an agent has more time for client conversations and follow-through. The benefit depends on verified accuracy, appropriate data use, and measurable results from the work.

Sources checked September 30, 2026. Product references describe the vendors’ announcements, not independent proof of performance. The listing example is hypothetical. The feature graphic is an AI-generated editorial illustration of a fictional property.

Stop Using AI Just to Write Listing Descriptions. Ask Why Your House Isn’t Selling.