Download websites that use Java

About Java

Web detection of a programming language is inference, and Java is the case where that has to be said before anything else. A language does not appear in a page the way a script library does. What appears instead is residue: a session cookie with a recognisable name, a path ending in an extension the runtime historically used, an application server identifying itself in a header, an error page with a familiar shape. Every one of those can be renamed, hidden or proxied away by a team that cares to do it.

The list is a floor, and a biased floor at that. Sites revealing a virtual machine underneath tend to be long lived deployments running close to default configuration, which is a fair description of enterprise infrastructure that has been in place for years. Newer deployments behind a reverse proxy and a modern framework can be entirely silent. The consequence is not that the data is unusable but that it describes a specific and commercially interesting subset rather than the whole.

What the subset implies is durable. A site with this runtime underneath was built by a team, budgeted by a company, and is maintained on a release schedule rather than on a whim. Banks, insurers, logistics operators, universities, government portals and industrial suppliers fill much of the population. Nobody stands up a virtual machine to publish a restaurant menu. Cost of ownership rules the technology out of the casual end of the web on its own, without anyone needing to decide against it. For anyone selling performance monitoring, migration consulting, security scanning, middleware or engineering recruitment, the qualifying signal and the buying signal are the same signal.

As of September 2026 we detect Java on 962,935 sites, with 588,720 of those in United States and .com as the recurring suffix. The language has been in use since 1995, which is longer than most of the sites running it have existed, and Oracle maintains the reference implementation under GPL v2. Longevity is the whole point. A technology that old on a public website is rarely a fresh decision, so each detection is a fact about accumulated history rather than about current taste, and the analysis worth doing is about persistence and replacement instead of adoption.

Persistence and replacement are precisely what a single lookup cannot show. Checking one domain tells you what one company runs today. The export lets you hold a population still and watch it move: which of these sites put a browser framework in front of the old application, which migrated onto a hosted commerce platform and retired the runtime completely, which added a balancer, which changed nothing at all for years. Migration on this scale is slow and visible only in aggregate across successive releases. Anyone whose business depends on legacy modernisation, or on knowing which organisations still have a legacy estate to modernise, needs the full set with its caveats stated openly rather than a confident number that quietly excludes everything hiding behind a proxy. A known undercount can be reasoned about, while an unknown one cannot.