Download websites that use Graphene

About Graphene

Some entries in a technology dataset are simple lookups. Graphene is not one of them, and a researcher is better served by understanding why before using the list.

Start with what the data actually contains. Across the detected population we separate 132 distinct version strings, and the value recorded most often is 2.9.4. A spread that wide is not what a single tightly released product usually looks like on the web. Fragmentation at that scale tends to mean either a component that ships bundled inside other software and therefore ages at the pace of whatever contains it, or a name that has been attached to unrelated projects over the years. The record here carries mixed signals of its own, since the classification sits under WordPress themes while the recorded implementation language is PHP, and the project name is shared across both of those ecosystems.

The honest conclusion is that the label identifies a fingerprint rather than a single vendor's product. Anyone building on the list should verify a sample of matched domains by hand and decide which population they actually care about before treating the set as homogeneous. Checking a handful of domains from the sample costs almost nothing and stops a campaign or a research claim from resting on an assumption. We would rather say so on the page than let a clean looking count imply a precision the detection does not have.

With that caveat in place, the list remains useful, and for some questions nothing else works. A component released under MIT in 2015 carries no billing relationship and no registration, so nobody, including the people who wrote it, holds a list of the sites using it. Usage can only be observed from the outside, by looking at what pages actually serve. Observation at that scale is the gap the dataset fills.

As of October 2026 the detection covers 1,808 sites, and the version column is where most of the value sits. Version distribution is what a dependency researcher needs to size the exposure created by a published advisory, what a maintainer needs in order to judge whether an old release can finally be dropped, and what anyone offering upgrade or modernization services needs in order to find the sites that are far behind rather than the ones already current. Sorting the population by version turns a flat domain list into a prioritized worklist.

The count is a floor in the usual way. Code that is minified, bundled or renamed during a build loses the markers detection depends on, and a version is only readable when the delivered asset still exposes one, so both the total and the version breakdown lean toward installations assembled in a simple and visible way. Sites with sophisticated build pipelines are underrepresented, which matters if the question is about modern applications rather than long lived ones.