Immowelt Scraper for German Property Listing Data

Not an open one. Immowelt documents a partner API built on SOAP web services exchanging XML over HTTP: a LocationService that turns a place name or postcode into a GeoID, an EstateService that returns lists, and an EstateExpose service that returns one listing by its GUID. Every call needs an API key issued by Immowelt, so it is a partnership integration rather than self-service signup. That documentation is version 3.1.0, dated 25 March 2015, and we read it from a Wayback Machine capture taken on 20 July 2024; it also shows how the interface behaves under load, since a broad list query returns ResultTooLarge and the caller has to narrow the criteria or subdivide the region. It further obliges the integrator to display an attribution note crediting immowelt.de. If you do not hold a partnership, collecting the public pages is the route, and that is what we do.

Read on 27 August 2026, the immowelt.de robots.txt disallows the parameterised endpoints - */classified-search?*, */classified-map?* and /classifiedList/ - along with map views such as /liste/karte and /expose/*/karte, preview exposes at /vorschau/expose/*, and the saved-search and booking areas. It blocks two named crawlers outright and gives AhrefsBot a Crawl-Delay of 50. The readable /suche/ paths and /expose/ pages are not in the disallow list, and the file publishes the sitemap index. We scope each project to publicly visible pages, honour those directives and pace requests. We do not bypass access controls, log-ins or bot protection. On our test date a plain HTTP client received a normal 200 on a search path and a 403 on an expose path, which is why detail collection runs through a real browser engine at a measured rate, and why a page that will not serve is marked rather than invented.

Only if you ask for it. Immowelt listings can carry a named private seller or a named contact at a Maklerbüro, so the provider block is the one place where output can touch personal data. Our default is listing and commercial data - price, size, rooms, location, features, dates, listing id and the provider's company name. Personal contact details can be excluded on request, and the provider block can be suppressed entirely. Under the GDPR you are the controller for whatever you choose to store, so the scope is worth deciding before the first run rather than after. We are not your legal advisers; we build the extract to the boundary you set.

Immowelt takes an expose down once the property is let or sold, and the UUID is what lets you notice. Each run records which ids were still present in the result pages, so a listing that stops appearing is dated rather than quietly dropped. From that you get time on market and a price change history per listing. We also flag likely relistings by matching municipality, Wohnflaeche, Zimmer and provider across runs, which catches a property that returns under a different id. Historical rows are kept rather than overwritten, so the series stays auditable.

Coverage is defined by the URL set, not by a quota. Immowelt prints its own result count in the page title, which gives you a check against our row counts: measured on 27 August 2026, the Berlin rental search read 6.808 Angebote, Munich 2.656 and Hamburg 3.058. Those figures move daily, so treat them as measurements with a date attached, not as constants. Depth is the part that needs care. On the readable /suche/ path a page parameter did not change what the server returned in our test, so we get depth by splitting geography instead - ad06de Kreis, then ad08de Gemeinde, then ad09de or nbh2de Stadtteil - until each slice is small enough to walk to the end.