Zillow Property Listings Scraper

One search in, one flat table out. Give it a location — a city, a ZIP, a county, or a Zillow search URL pasted from your browser — and it returns the properties Zillow would show you, each with its address, price, beds, baths, floor area, lot size, home type, days on Zillow, coordinates, listing broker and photo.

Covers all three Zillow tabs: for sale, for rent and recently sold.

Built to be called by an agent or a script: small input, small output, one row per property, no nested objects, no HTML.

Input

field	meaning
location	"Austin, TX", "90210", "Miami-Dade County, FL", or a full Zillow search URL
listingType	forSale (default), forRent or sold
maxResults	How many rows to return. Default 100, ceiling 820 — see below
priceMin, priceMax	Price band. Asking price for sale and sold; monthly rent for rent
bedsMin, bathsMin	Minimum bedrooms and bathrooms
sort	Zillow default, newest, price, bedrooms, square feet or lot size
requestIntervalSecs	Pause between result pages. Default 1
proxyConfiguration	Residential proxy. Required in practice — see below

Pasting a Zillow URL keeps whatever filters are already on it — waterfront, school rating, keyword, HOA — and applies the fields above on top. The one thing a pasted URL does not override is listingType, because an input that loses to a value hidden inside a URL is an input nobody can trust.

Output

One item per property:

field	meaning
zpid	Zillow's property id. Stable for a home; a lat--lng string for an apartment building
url	The listing page on zillow.com, absolute
address	Full address as Zillow renders it
street_address, city, state, zipcode	The same address, split
price	Asking price, monthly rent, or sale price, depending on listingType
price_text	The same figure as Zillow renders it, e.g. "$3,455+"
currency	Currency price is quoted in
beds, baths	Bedroom and bathroom counts. Zillow counts a half bath as 0.5
living_area, living_area_unit	Interior floor area, sqft for US listings
lot_size, lot_size_unit	Lot size, and the unit it was measured in — read both
home_type	SINGLE_FAMILY, CONDO, TOWNHOUSE, MULTI_FAMILY, APARTMENT, LOT, …
listing_status	FOR_SALE, FOR_RENT, SOLD, PENDING, …
days_on_zillow	Days the listing has been live
latitude, longitude	Coordinates
broker_name	Listing brokerage
image_url	First photo, full size
zestimate, rent_zestimate	Zillow's own value and rent estimates
year_built	Always null — see below
is_building	true for an apartment building rather than a single home
listing_type, search_location	The search this row belongs to, echoed back

lot_size needs lot_size_unit

Zillow switches units by size: a suburban lot comes back as 6359.76 sqft and a rural one as 2.4 acres. A pipeline that reads lot_size alone and sorts on it will put the two-acre parcel below the townhouse. The unit always travels with the number.

year_built is declared but never filled

Zillow's search endpoint does not carry it. The value lives only on each property's own detail page, which would be one extra request per row — a different, and much more expensive, product than this one. The column is kept so this dataset joins cleanly onto detail-page data; it is never guessed at.

Apartment buildings are nulls, not zeros

A for-rent search returns individual homes and whole apartment buildings. A building has no single price, no bed count and no floor area, because it is a container for units that each have their own. Those rows carry is_building: true, null in the per-home fields, and the cheapest advertised unit price in price with Zillow's own "$3,455+" string in price_text.

A 0 there would read as "studio, no space, free", and any screen built on the dataset would pick it up. Nothing in this output is ever zero to mean "unknown".

Limits worth knowing before you run it

• 820 rows per search, always. Zillow serves 41 listings per page and stops paginating at 20 pages, whatever the match count. total_available in the run log tells you how many matched — it is routinely five figures. To collect more than 820, split the search: by ZIP, by price band, or by home type.
• Residential proxy is required, not optional. Zillow refuses datacenter addresses, and is uneven even on residential ones — the same page answers HTTP 200 from one exit address and 403 from the next, minutes apart. The run answers a refusal by rotating to a fresh address and backing off with jitter rather than failing, which is what turns that unevenness into data. Turning the proxy off will produce blocks.
• Zillow's anti-bot page is an HTTP 200. It serves a "Press & Hold" page rather than an error status, so a status check alone would report success and return nothing. This actor treats a page with no data in it as a refusal and rotates, which is why a run can log retries and still come back complete.

Pricing

Pay per event: a small charge when the run starts, plus one charge per listing row delivered. Rows skipped as duplicates across page boundaries are not charged, and a run that returns nothing charges only the start event.