Real Estate Listing Extractor (single page)

Turn a single public real-estate listing page URL into a clean, structured JSON record — deterministically, with no AI, no proxy, and no headless browser.

What it does

Given one listing-page URL (or a small bounded batch), the actor fetches the page once and extracts structured listing data from its embedded schema.org markup. It is built on the same proven, SSRF-guarded fetch core as the other OMEGA single-page extractors, with a deterministic real-estate parser on top. Pure code — every field is computed, never guessed by a language model.

Input

Field	Type	Description
url	string	A single public real-estate listing page URL (include https://).
urls	array	Optional bounded list of extra listing URLs (max 50 per run).

Example input:

{
  "url": "https://www.example-realty.com/listing/123-maple-st"
}

Output

One dataset item per URL:

{
  "url": "https://www.example-realty.com/listing/123-maple-st",
  "status": "completed",
  "address": "123 Maple St, Austin, TX, 78701, US",
  "price": "675000",
  "currency": "USD",
  "beds": "4",
  "baths": "2.5",
  "area_sqft": "2400",
  "property_type": "singlefamilyresidence",
  "listing_type": "sale",
  "year_built": "1998",
  "lot_size": "6997",
  "agent": "Acme Realty",
  "images": ["https://.../a.jpg", "https://.../b.jpg"],
  "description": "Charming family home with garden.",
  "geo": { "lat": "30.2672", "lng": "-97.7431" },
  "raw_prices": ["675000"],
  "method": "jsonld_realestate",
  "parse_confidence": "high",
  "extracted_at": "2026-06-24T10:00:00+00:00",
  "error": null
}

status is one of completed, failed, blocked, or empty. Any field that the page does not declare is returned as null (or []) — never invented.

Use cases

• Normalise a listing URL into a row for a CRM, spreadsheet, or database.
• Pull price / beds / baths / area for a comparables (comps) sheet.
• Monitor a single listing's price and status over time.
• Enrich an internal dataset of listing URLs with structured fields.

How it works

Extraction precedence (most reliable first); the layer used is reported in method:

1. schema.org JSON-LD — RealEstateListing, Residence (House, Apartment, SingleFamilyResidence, …), Accommodation/Place, and Product/Offer (price, currency, address, bedrooms, bathrooms, floor/lot size, year built, geo, images, broker/seller, sale-vs-rent via businessFunction).
2. OpenGraph / product meta — og:title, og:image, product:price:amount, product:price:currency, location meta.
3. Meta / heuristics — <title>/<h1> plus a conservative, currency-marked price detector (never infers a price from a bare number).

Areas declared in square metres (unitCode MTK) are converted to square feet. A code-owned parse_confidence (high/medium/low/none) reflects which layer matched and how many core fields were found.

Cost-safety

• No proxy, no headless browser, no LLM, no paid API. One bounded HTTP GET per URL (hard caps: 5s connect / 10s read / 2 MB / 3 redirects).
• $0 idle and $0 uncovered cost beyond Apify compute — nothing to subsidise.
• SSRF-guarded and fail-closed: private/loopback/reserved IPs are blocked, with per-redirect re-validation, and a domain blocklist for bot-walled portals.

Limitations (honest)

• This is single-page extraction, not bulk portal/MLS scraping. It fetches the page you give it and never follows links or paginates.
• It only reads server-rendered markup. Pages that render entirely in the browser (heavy client-side JS) will expose little to a plain GET and return a low-confidence result.
• Many large portals (Zillow, Realtor.com, Redfin, Rightmove, Zoopla, …) block bots and/or forbid scraping in their ToS — these are on a blocklist and return status: "blocked". Point the actor at a brokerage's or publisher's own listing page that exposes schema.org markup for best results.
• Fields are only as good as the page's structured data. Missing data is returned as null; the actor never fabricates a value.