Download websites that use Python

About Python

Server-side languages are inferred rather than observed. Nothing in a page tells you what executed to produce it unless the machine volunteers the information, in a response header, an error page, a session cookie name, a framework specific URL pattern or the layout of static asset paths. Every count of Python on the public web, ours included, is a floor, and it is biased toward servers that talk about themselves.

The version column makes the bias visible and is worth reading carefully before anyone quotes it. The release we record most often is 2.7.5, a branch that stopped receiving upstream maintenance long ago, out of 225 distinct versions in total. The honest reading is not that the modern web runs on an obsolete interpreter, but that the machines which announce their interpreter tend to be the ones running an operating system package on a long support cycle, while a container built this year says nothing at all.

A second bias runs alongside the first. Frameworks differ in how loudly they identify themselves, and a stack fronted by a reverse proxy that strips headers will hide its language entirely while an identical application behind a default configuration announces it. The visible population is therefore shaped as much by operational habits as by technology choice, and any comparison between countries or hosting providers is partly a comparison of how their servers are configured.

What presence implies commercially is stronger than what it implies technically. A site with this language behind it has an application, which means an engineering team or a contractor, a deployment process, and usually adjacency to data work, since the same language carries analytics, machine learning and internal tooling. Domains where the language is visible are prospects for hosting, observability, error tracking, dependency scanning and developer recruitment, and they are a very different population from the sites assembled in a page builder.

As of September 2026 we record it on 865,055 sites, 488,788 of them in United States, with .com the domain ending we see attached most often. Released in 1991 and distributed as Open Source, the language has had a long time to spread into places where nobody thinks of it as a web technology at all, which is part of why the visible footprint and the real footprint diverge so widely.

The gap between those two footprints is the reason to work from a list rather than a lookup. A checker tells you whether one domain revealed a language on the day you asked. A file with one row per domain lets you measure how often the language is revealed at all, compare that rate across countries and hosting providers, isolate the sites that leak version strings and treat everything else as unknown rather than empty. Absence in this dataset should be read as not observed, never as not used, and that distinction only becomes usable when the whole population is in front of you instead of a single answer.

Used that way the file supports the questions that matter for sizing a developer market. Which hosts concentrate visible deployments, which country segments expose more than others, how the version spread moves between snapshots, and which of your own accounts sit in the group that leaks enough to be tracked at all.