Download websites that use Google Cloud CDN

About Google Cloud CDN

A content delivery record is a statement about the path a request took, not about where the site actually lives. The origin can be anywhere, run by anyone, on any stack, and the edge layer in front of it is chosen separately and changed more often. Rows of this kind therefore sit at a different level of a technology profile than hosting rows, and they can coexist with any of them, which is why treating delivery and hosting as one question produces confused analysis.

Evidence for delivery is comparatively firm. Cache status headers, edge identifiers and the shape of a response are functional rather than cosmetic, so they are less prone to the guesswork surrounding library and framework fingerprints. Firm is not the same as complete, though, because headers can be stripped or overridden at the edge, some configurations answer only for asset hostnames, and a domain can serve its own homepage directly while pushing images and scripts through an entirely different network.

Something specific about Google Cloud CDN shapes how the population should be read. The service is bought as part of the wider platform from Google rather than as a standalone contract, normally attached to the same load balancing layer, so its footprint tracks depth of commitment to one supplier rather than an independent shopping decision. Available since 2015 and sold as Proprietary software, the service rarely turns up on domains with no other relationship to the same vendor, and a positive record is usually a clue that mail, compute or analytics sit there too.

As of September 2026 we see it on 8,913,953 sites, 6,699,408 of them in United States, with the list skewing toward .com. Geography on an edge network row reflects the registration country of the domain rather than where traffic is actually served, since an edge network exists precisely to serve everywhere at once, and reading the column as a traffic map inverts what it measures.

Buyers of delivery data are a distinct group. Competing networks want migration targets and a sense of renewal timing. Performance and monitoring vendors want sites whose delivery layer is already sophisticated enough to justify a conversation. Security teams study which parts of the web depend on which edge providers, because correlated outages and correlated policy decisions are a systemic risk that only becomes visible at population scale. Researchers working on web centralisation want the same counts for the opposite reason.

Checking one hostname takes a single request, and any engineer can do it before lunch. What cannot be improvised is a filtered population, where the delivery signal is combined with country, suffix, platform and origin to answer whether a segment is winnable, how much of it already runs more than one network, and which rows changed since the previous snapshot. Expect the file to under report sites using several networks at once, and to describe the hostname we crawled rather than every hostname a company owns. Stated plainly, the download answers population questions about the edge of the web, and it leaves the questions about any one company to a request you can make yourself.