Download websites that use Hugo

About Hugo

Hugo is an awkward case for anyone who builds technology datasets, because a static site generator is supposed to disappear. By the time a visitor loads the page, the Go binary that produced it has finished running on somebody's laptop or inside a build container, and what remains on the server is plain HTML that could have come from anywhere. Detection rests entirely on the traces the generator chooses to leave behind, mostly a generator meta tag and the file and theme conventions that the default templates emit.

Any figure built on that basis is an undercount, and we would rather say so than present it as a census. Removing the meta tag is a single line of configuration, and teams who care about not advertising their stack remove it. The bias runs in a predictable direction: our population leans toward default configurations, documentation sites and personal or team blogs, and leans away from the agencies and product companies who customise the build. As of September 2026 we detect 176,692 sites, which should be read as a lower bound on a technology that is deliberately quiet.

Where the trace does survive, it is unusually informative. A site built this way has no database, no admin login and no server-side runtime to attack, which removes a whole class of security and hosting spend from the conversation. Somebody chose a Static site generator that requires editing Markdown in a repository and running a build, so there is a developer in the loop, on staff or on retainer. Marketing tools that assume a plugin directory simply do not apply. A vendor selling managed PHP hosting, plugin security or a page builder has nothing to sell to this audience, while a vendor selling continuous integration, edge hosting or a headless commerce API is talking to people who already made the architectural decision that the pitch depends on.

Version data is where the dataset earns its keep on this technology. We have observed 390 distinct releases in the wild, and the build most often present is 0.147.7, which is a wide spread for a tool with a fast release cadence and no automatic update path. A static site freezes the generator at the moment of its last deploy, so the version string is really a timestamp. Reading it tells you when the site was last built, not what its maintainers run today, and a long tail of old releases is a long tail of projects nobody has touched. Separating those from actively maintained sites is a filter that no per-domain lookup can give you, because the comparison only exists across the whole population.

The concentration sits in United States, where 122,775 of the detections live, with .com running through the rest, which reflects where the developer audience for a Apache 2.0 project first formed after its release in 2013. For a researcher the useful move is to pull the entire list rather than test domains one at a time, since the questions worth asking are aggregate ones: how the static generator share of a market has changed, which hosting providers those builds land on, and which sites have not been rebuilt since a release several generations back.