The agent tool landscape
What a Zestimate is doing under the hood, and the inputs it cannot get
6 min read
A Zestimate reads records and comparable sales. Its accuracy is set by what it cannot read.
The short version
A Zestimate is an automated valuation model that estimates a home value from public records, tax data, prior sales, and comparable nearby sales, refreshed on a schedule. On listed homes it looks accurate. On off-market homes its median error runs near 7.5%, roughly $45,000 on a $600,000 home, because the inputs that move a real sale price the most, condition, updates, layout, and the specific block, are the inputs a model cannot get. Knowing what it sees and what it is blind to is how you read the number correctly.
What the model runs on
A Zestimate is not a person looking at the home. It is a model reading data. Its inputs are the structural facts in public records, beds, baths, square footage, lot size, year built, the tax assessment history, the record of prior sales, comparable recent sales in the area, and, where a homeowner has submitted them, owner-supplied facts. From those it produces a single figure and refreshes it on a schedule. Everything it knows about a specific home is on that list.
What it cannot see
The problem is that the inputs that move a sale price the most are not on that list. The table below separates what the model can capture from what it cannot.
| Signal that moves price | Can the model capture it? | Why it matters |
|---|---|---|
| Beds, baths, square footage, lot, year built | Yes, from public records | The structural facts a model prices well when the records are current |
| Recent comparable sales nearby | Yes, where sales data exists | The main driver of the estimate, and the reason thin-data areas break it |
| Interior condition and finish quality | No | A dated interior and a renovated one can differ by tens of thousands at the same square footage |
| Recent renovations not in records | No | A new kitchen or roof rarely shows up until a sale, so the model misses the value it added |
| Layout and flow | No | Two homes with identical square footage can sell far apart if one flows well and one does not |
| View, exposure, and street-level facts | No | A busy road, a backing power line, or a rare view moves price and is invisible to records |
| Off-market home with stale records | Poorly | With no recent MLS signal the model leans on old data, which is where its error is widest |
Why off-market homes break it
The estimate is at its best on a home that recently sold or is listed, because fresh, similar sales feed it directly. It is at its worst on an off-market home in a thin market, where there is no recent MLS signal and the model falls back on old records and distant comparables. This is why the published median error for off-market homes runs near 7.5%, about $45,000 on a $600,000 home, and median means half of all estimates miss by even more. The number is not random, but it is a wide guess dressed as a precise figure.
The trailing-digits illusion
When the tool reports a value of $612,438, the trailing digits read as certainty, and that is the trap. A figure to the dollar looks far more precise than a model carrying a $45,000 band of error can support. The honest version of that number is a range with a reason, likely between one figure and another, wider where the data is thin. A number shown alone hides its own uncertainty, which is exactly what makes it feel more trustworthy than it is.
What this means for an agent
The move at a listing appointment is not to tell the seller the Zestimate is wrong. It is to show what it could not see on this home. Walk the specific inputs it missed, the renovation that never hit records, the layout, the corner lot, then set the price on recent comparable sales and the condition you can see with your own eyes. That reframes the conversation from the algorithm versus you to the record versus the reality of this house. For the fuller picture of why the estimate misses, the consumer explainer why the Zestimate is wrong on my house is written for the client to read, and on the mechanics see how AVMs work and where they break. Why a single number should carry a range at all is the subject of what confidence should mean in a valuation tool. A reasoned estimate that shows its range instead of a false-precise figure, like our home-value page, is built on this exact gap.
Frequently asked questions
What data does a Zestimate use?
It uses public records such as beds, baths, square footage, lot size, and year built, along with tax assessment history, prior sales, comparable recent sales nearby, and any owner-submitted facts. It does not use a physical inspection of the home.
Why is the Zestimate wrong on my house?
Usually because the inputs that set your home apart are not in the data. A model cannot see condition, a renovation that never reached public records, layout, or street-level factors, so a home that is much better or worse than its records suggest will be mispriced.
Can a Zestimate see the inside of my home?
No. It has no view of interior condition, finish quality, or layout. It prices the home from records and comparable sales, which is why two homes with the same square footage but very different interiors can receive similar estimates.
How accurate is the Zestimate?
It varies by home. Zillow reports a median error near 7.5% for off-market homes, which is about $45,000 on a $600,000 home, and it is generally tighter on listed homes and wider on unique or thin-data ones. Median error also means half of estimates miss by more than that.
Does updating my home facts on Zillow change the Zestimate?
It can, because owner-submitted facts are one of the model inputs, but corrected square footage or bed and bath counts move the number more than a note about finishes. The model still cannot see condition, so updates help most when they fix a factual error in the record.
Sources: Zillow published Zestimate methodology and accuracy disclosures, including the off-market median error figure, general descriptions of automated valuation model inputs, and standard real estate valuation references on the drivers of sale price. Figures vary by home and market; treat them as general references.
This article is general information and analysis, not financial, lending, or appraisal advice. Verify any home value with a licensed professional before acting.
The Independent Agent
Substack | Spotify | CMAflow FAQ | YouTube | Free CMA | Home valuation | Insights | Blog