A Zestimate is a starting point, not a price. Zillow’s own published accuracy is a median error of about 1.9% for homes already on the market and about 7.0% for homes that are not, as of August 2026, and median means half of homes miss by more than that. In Chicago, the misses concentrate in exactly the homes this city is made of: condos, two-flats, and older housing stock where the algorithm cannot see the things a buyer prices first.
I list homes across Chicago and the suburbs, and the Zestimate comes up at almost every kitchen table. This page is the full answer: what the number actually is, where it goes wrong here, what the error rate means in dollars, and how to get to the real number before you commit to a list price.
A real example of the gap
A recent one from my own pipeline, details changed to protect the seller. The Zestimate said $1.62 million. The defensible number from recently sold comparables was $1.475 million. That gap is $145,000 on one home. This is a real example, not a promise about your home, and the direction is not always high. The point is the size: an automated estimate can miss by more than most people’s annual income, in either direction, and it reports that guess with a straight face.
Overpricing off a high anchor is not a free test, either. In the suburb data I publish, overpriced listings sat weeks longer and still cut their price. The anchor you start from shapes the outcome you get.
What a Zestimate actually is
A Zestimate is an automated valuation. The algorithm reads tax records, past sales, comparable sold homes, market conditions, and, importantly, active listing prices in the area, then produces a number for nearly every home in the country, including homes that have not sold in decades and homes the model has never seen inside.
That last part is the entire story. The model has never been inside your home. Everything it knows is what is on paper.
The four things the algorithm cannot see in Chicago
These are the Chicago-specific factors I watch a Zestimate miss, over and over.
1. Floor level and garden units. A garden or below-grade unit prices differently than the same square footage upstairs. Another real example from my pipeline, details changed: the automated range on a garden unit said $345,000 to $362,000. The defensible sold-comp number was about $295,000 to $315,000. The algorithm cannot see that. A buyer can.
2. Whether a bedroom is legally conforming. In Chicago housing stock, a “third bedroom” with no closet or no egress window is not a third bedroom to an appraiser or a lender. The tax record says 3. The market pays for 2. The algorithm cannot see that. A buyer can.
3. Parking. Deeded spot, leased spot, permit street parking, or nothing changes real value by tens of thousands of dollars in many neighborhoods, and it is inconsistently recorded. The algorithm cannot see that. A buyer can.
4. Thin comp pools. Automated estimates are most confident where hundreds of similar homes trade constantly. A vintage two-flat in a neighborhood where only four have sold this year gives the model almost nothing to work with, so it borrows from homes that are not really comparable. The algorithm cannot see the difference. A buyer can.
Condition sits underneath all four. Updated versus original kitchen, deferred maintenance, a roof at end of life: all invisible on paper, all priced by every buyer who walks through.
What “median error” means in dollars
Zillow’s published numbers, as of August 2026: about 1.9% median error for on-market homes, about 7.0% for off-market homes. Two things about those figures deserve plain arithmetic.
| The stat | What it means on a $500,000 Chicago home |
|---|---|
| 7.0% median error, off-market | $35,000, and half of off-market homes miss by MORE |
| 1.9% median error, on-market | The model gets sharply better after a list price exists, because it leans on the list price |
Read the second row again. The estimate is most accurate for homes that already have a human-chosen price to anchor on. When you are deciding what to list for, you are in the first row, the 7% row, by definition. The moment you need the number most is the moment it is least reliable.
What I actually use the Zestimate for
Honestly: I look at it before almost every listing appointment. Not to price the home. To read the room. The seller has usually been watching that number for months, and it tells me what expectation I am walking into. When the Zestimate and the sold comps agree, pricing conversations are easy. When they are $145,000 apart, the first job is not marketing, it is evidence.
That is the fair way to treat it. A Zestimate is a decent first guess and a terrible final answer. The people who get hurt by it are not the ones who glance at it, they are the ones who anchor on it.
What actually sets the number
A buyer sets your home’s value, and a buyer prices three things the estimate cannot weigh for your specific block and your specific home: demand (how many buyers are competing in your price band), supply (how many homes they can choose from), and speed (how fast comparable homes are going under contract). Recently sold comparables are the record of what real buyers actually paid for homes like yours. That is why an appraiser starts there, and why I believe a list price should too.
How to get the real number, free
Three steps, and the first two cost nothing and work whether you ever contact anyone.
- Fix your home facts on Zillow. Wrong square footage, bed or bath counts, and missing renovations degrade the estimate. Zillow says data quality directly affects the number. Correct it for accuracy, just know you are improving a guess, not replacing evidence.
- Pull your own comps. The last 60 to 90 days of sold prices for your property type, your size, your condition, your area. Three to six good matches beat any algorithm. While you are there, count how many active listings near you have already cut their price. That tells you what the anchor is doing to your neighbors.
- Have the MLS pull run for you. I run this as a free neighborhood report at Net Gain Realty: sold comps, demand, supply, days to contract, and the overpricing penalty in your area, from the same MLS data an appraiser will use. No obligation, and you can take it to any broker you want. To see what any price nets you after real Chicago closing costs, use the home sale calculator.
The reason I care this much about a free estimate is simple. The list price decision moves more of your money than any other choice in the sale except one, the listing fee itself. Both deserve evidence instead of a default.