Download websites that use Webpack

About Webpack

Almost everything else on a technology page describes something that runs in a browser or on a server. Webpack runs on a developer's machine, finishes its work long before anyone visits the site, and is then not present at all. What we detect is residue.

The bundler takes an application's source files and their dependencies and emits a compact set of files fit for a browser. Along the way it leaves recognizable shapes behind: the runtime helper that fetches code chunks on demand, the naming convention applied to those chunks, and the module registry structure wrapped around the compiled output. Released in 2012 under the MIT license, the tool standardized those shapes across much of the ecosystem, which is exactly why they can be read from outside. We publish no version for it, because a compiled bundle does not announce one and inferring a release from code shape would be a guess dressed up as a fact.

As of September 2026 the residue is visible on 2,471,039 sites, filed under Miscellaneous, with United States accounting for 1,559,103 and .com carrying much of the rest. Distribution here tracks where application-style front ends get built rather than where the web is, which is a genuinely different map from the one a content management system would draw.

Presence carries an inference that matters commercially. A site with bundler residue was compiled, and a compiled site has a package manifest, a dependency tree, a toolchain somewhere in its history, and a person or a team paid to maintain all of it. Brochure pages assembled in a builder do not look like this. Whoever sells developer tooling, error monitoring, dependency scanning, or engineering recruitment is looking for precisely the population a bundler fingerprint isolates, and the fingerprint holds that population steady without depending on which framework happens to be fashionable in a given year.

Security research uses the same signal from the other side. Bundled output frequently ships with source maps left in place, dependency versions embedded in compiled code, and internal module paths preserved inside chunk names, so the population is also the population worth checking for accidental disclosure. A list of compiled sites is a reasonable starting frame for that kind of study, and there is no realistic way to assemble one by hand.

The undercount is severe and no phrasing softens it. A build can be configured to erase every marker we look for, and meta-frameworks that wrap the bundler often emit output carrying their own signature instead of the underlying one. A site can compile with the tool and look, from outside, as though it never did. Absence is close to meaningless here, while presence is reliable, and that combination is the right shape for prospecting and the wrong shape for market share arithmetic. Anyone quoting an ecosystem percentage from this column is quoting a floor.

A per-domain lookup fits none of this. Checking a single site for bundler residue is a devtools exercise anyone can do in a minute, and the answer supports no decision. The downloadable list answers what a single check cannot: how large the compiled web is, how it distributes across countries and domain suffixes, which frameworks and monitoring tools travel alongside it in the same records, and where a segment of engineering-led sites can be found and sized before anybody spends money trying to reach it.