The robot does one thing: it looks at homes that sold nearby, takes their price per square foot, and multiplies by this home's size. Basically "what do homes round here go for, times how big this one is."
That works great for a normal home on a normal street — the neighbors really are similar. That's why the Miami house came out perfect.
But the robot is blind to why a home is cheap: the Boynton condo sold for $75k because it's a small, age-restricted unit with a big monthly fee — the robot can't see that, so it priced it like the normal condos next door and came in too high.
And it's blind to why a home is fancy: the Jupiter home sold for $3.3M because of the water, the view, the dock, the finishes — the robot can't see that either, so it averaged the smaller inland homes nearby and came in too low.
This is not "the valuation is wrong." It's the opposite. You walk the property. You see the water, the condition, the club fee, the reason. The robot can't.
So on plain-vanilla homes the robot is a fine gut-check — but on anything special, your eyes beat it every time. The study doesn't say your judgment is off. It says exactly which homes still need a human, and that human is you.