You want a machine-readable dataset rather than a website, and the aggregator has already done some of the work for you.

Scraping all of Zillow's sources instead would explode the complexity of the project.