Americans love Zestimates . When Zillow debuted in 2006, the fledgling site bore little resemblance to the real-estate behemoth it is now. Then the search portal couldn't even tell you which homes were actually for sale.

There was, however, the Zestimate: a "free, unbiased valuation" for 40 million houses around the US, based on a proprietary algorithm.

The public reaction was immediate and frenzied. Curious homeowners and nosy neighbors crashed the site within hours of its launch. Suddenly, half of the single-family homes in America had a dollar figure attached, and everyone wanted a peek. It was like the housing market had gone full reality TV - transparent, a little scandalous, and impossible to ignore.

Fast-forward to today, and the Zestimate has become a household name. For many, it’s a convenient way to check on the health of the housing market. Is your home up 20% in value? Have mortgage rates budged? The Zestimate has you covered - or so people think.

But for real estate insiders, the Zestimate isn’t exactly a trusted oracle. At best, it’s a decent starting point. At worst, it’s a digital blindfold leading homeowners astray.

So, what is a Zestimate, really? It’s one of many Automated Valuation Models (AVMs), the digital tools banks, investors, and homeowners use to estimate property values. Zillow’s secret sauce includes thousands of data points, public records, tax documents, and multiple listing services (MLS). Sprinkle in user-submitted details, like your newly renovated kitchen or backyard pool, and voilà: a dollar figure for your home.

Zillow claims a median error rate of 2.4% for homes actively on the market and 7.49% for off-market homes. On paper, those numbers look impressive. For on-market properties, it suggests that half of all Zestimates fall within 2.4% of the eventual sale price. But for the other half? Things get murky. Zillow doesn’t say much about the outliers, and the accuracy of off-market estimates is even shakier.

Here’s the kicker: Zestimates get significantly better when a house hits the market. That’s because human intervention ... via real estate agents, listing prices, and detailed photos feeds the algorithm critical context it didn’t have before.

When a home is listed, the Zestimate adjusts, moving closer to the seller’s asking price. From there, it continues to refine itself based on market activity until the house is sold. At that point, Zillow uses the final sale price to calculate its on-market error rate. Essentially, the Zestimate becomes smarter by piggybacking on the real work done by humans.

Without a listing price to anchor its calculations, the Zestimate is flying blind. The moment a seller puts a number on their home, the Zestimate snaps into focus. And because most sellers rely on a real estate agent, that list price is often grounded in local market knowledge, comparable sales, and the unique quirks of the property. In September 2024, Zillow reported that the typical home sold for 99.8% of its list price, almost a perfect match.

For Zillow’s algorithm, the list price is pure gold. In the world of AVMs, models that gravitate toward list prices are nicknamed “springy” or “bouncy,” like a ball tethered to a string. They’ll stretch a little but rarely stray far from their anchor point.

Real estate professionals have noticed a peculiar phenomenon: when a seller lists their home with a price significantly different from the Zestimate, the Zestimate often shifts to align with the list price. It’s as if the algorithm is taking cues from the seller rather than standing firm on its original valuation.

Zillow doesn’t hide this. In fact, the company openly admits that list prices heavily influence Zestimates. It’s a pragmatic move list prices are, after all, informed by human expertise.

Here’s the thing: maybe accuracy isn’t the point. Back in 2006, when the Zestimate launched in beta, Zillow wasn’t aiming for perfection. “The Zestimate started out fairly inaccurate, but it didn’t matter,” Rich Barton, Zillow’s cofounder, said in a 2021 podcast. “It was provocative.”

And provocative it was. The Zestimate turned an opaque housing market into a spectacle, giving everyone from first-time buyers to seasoned investors a reason to log on. Even when the numbers were wildly off, the Zestimate sparked conversation, curiosity, and controversy.

Zillow’s former CEO, Spencer Rascoff, once experienced this firsthand. In 2016, he sold his own home for 40% less than its Zestimate. Did he throw the algorithm out the window? Not quite.

In fact, the following year, Zillow doubled down, offering $1 million to anyone who could significantly improve the Zestimate’s accuracy. A trio of data scientists from the U.S., Canada, and Morocco rose to the challenge, boosting the algorithm’s performance by 13%.

So, where does this leave us? For all its flaws, the Zestimate has undeniable value as a conversation starter. It gives homeowners a rough idea of their property’s worth and sparks interest in the housing market. But as a precise valuation tool, it falls short especially for homes not actively listed for sale.

If you’re buying or selling a home, treat the Zestimate as you would a weather app: a useful snapshot, but not gospel truth. Consult a real estate professional for the full picture. After all, no algorithm (no matter how sophisticated) can replace the boots-on-the-ground knowledge of someone who’s walked through the house, seen the neighborhood, and studied the market trends firsthand.

In the end, the Zestimate is a lot like the housing market itself: fascinating, frustrating, and never quite what you expect. Whether you love it or loathe it, one thing’s for sure - it’s here to stay.