Putting AI to the test with real estate pricing

Photo: Pixabay

Artificial intelligence may or may not be useful for pricing property, depending on the program one uses.

In my last few columns, I wrote about artificial intelligence or “bots,” and their uses for real estate selling, buying etc.

I looked at Grok from X , ChatGPT and Meta’s AI. There are other AI programs out there like Claude, Gemini, DeepSeek, Perplexity, Alexa, Cortana and more.

I decided to stay with what I consider to be the most popular but feel free to take a look at these other ones if they interest you.

This week, I asked the bots to determine a fair market value for one of my listings, in Kelowna. I put in the actual address and let them do their thing.

First up was Grok 3. It went all out and came up with 625 words which I asked to be trimmed down to 150 words.

“To price the home at (the address in Kelowna) I analyzed its 2,527-square-feet size, five bedrooms, 3.5 bathrooms and 9,583-square-foot lot in Black Mountain. Comparables on (the same road) show smaller homes (1,369 square-feet to 1,853 square-feet) valued at $854,100 to $908,800 in 2023–24. Kelowna’s January 2025 benchmark price for single-family homes is $1,030,600, up 5.4% year-over-year, but Black Mountain is slightly below average. Using a $490-square-foot to $550-square-foot range from comps, the home’s value is $1,238,230 to $1,389,850. High inventory (nine months as of September 2024) indicates a buyer’s market, so pricing competitively is key. The large lot and family-friendly features add value but an older home may need updates. I recommend listing at $1,199,000, balancing size, location, and market conditions. If renovated, $1,250,000 could work, but adjust if it sits. Consult a local realtor for precision.”

Its worth noting the home is in north Rutland, not Black Mountain, so Grok seems to be a little location challenged. I think the next version may need a GPS built in.

Stepping up to the plate next was Meta’s AI

“To provide a more accurate pricing suggestion (for the home), I'd need more information about the property. Here are some questions to consider:

What type of property is it (single-family home, townhouse, condo)?

What is the square footage of the property?

How many bedrooms and bathrooms does it have?

What is the age and condition of the property?

Are there any notable features (views, pool, etc.)?

What is the current state of the local real estate market?

You can also check websites like:

B.C. Assessment

Zestimate (Zillow)

These resources can give you an estimate based on recent sales data and property characteristics.

For a more precise valuation, consider consulting a local real estate agent.”

So, basically it just said I don’t know, ask someone else. But it took 134 words to do it. That sounds like some politicians you might know. A little tongue in cheek I know, however, the other AI programs did find out more about the property on their own and Meta’s AI didn’t try hard at all.

Last up to bat was ChatGPT, which took 334 words and the response paired down to 150 words was:

“Nearby comparable properties range from $789,000 to $1,125,000. For example, (a nearby home) (1,347 square-feet) is listed at $789,000, and (another nearby home) (2,452 square-feet) is priced at $900,000. Larger homes with suites or corner lots push closer to $1.1 million. Based on size, updates, and location, an estimated value between $900,000 and $950,000 is reasonable.

For an accurate price, a local real estate agent can provide a comparative market analysis factoring in renovations, lot utility and market timing. This helps position the home competitively for a successful sale.”

The answers ranged from $1.25 million to I have no idea to between 900,000 to 950,000. ChatGPT seemed to be in the ballpark, relying on list prices of nearby properties.

List prices, as you know, are usually determined by real estate agents and their clients, so ultimately it derived its price from prices people set. If this were a school assignment, I think I might say that is using someone else’s work or cheating.

In the end the market determines the price. When someone agrees to buy a property at a certain price, until then a true value hasn’t been set. As agents, we aim to price listings in the ballpark so the market can view the property and decide its value.

In today's market, with more inventory than buyers, overpriced homes can sit for a long time, waiting for that one perfect buyer. If a quicker sale is needed, pricing lower attracts more buyers who are often more flexible but expect a better deal. That is why an agent is valuable, not just to assess the property but also to understand the seller’s priorities and work with the client to achieve their goals.

I am going to say it was another swing and a miss for AI this time.

Score: AI zero and humans two.

If you have suggestions for other real estate-related columns, please email me at [email protected]

This article is written by or on behalf of an outsourced columnist and does not necessarily reflect the views of Castanet.