Download websites that use Apache HTTP Server

About Apache HTTP Server

Apache is old enough that the interesting question is no longer why it became dominant but why it has stayed. It has been serving pages since 1995, which makes it older than the commercial web that grew on top of it, and as of September 2026 we still detect it on 40,097,945 sites.

Software that large does not persist by being chosen. It persists by being already there. Shared hosting estates, control panel defaults, appliance images and internal tooling all carry it forward without anyone making a fresh decision, and each of those channels renews the population quietly.

The country split is where that shows up most clearly, and it is the number on this page that most often surprises people. Germany alone accounts for 13,570,874 of the sites we detect it on, a concentration well above what that country's overall share of the web would predict. There is no single explanation. Hosting market structure, the longevity of shared hosting there, and the installed base of control panels that ship it by default all plausibly contribute. What matters for anyone using the data is that the pattern is stable across rebuilds rather than an artefact of one crawl, so it can be planned around.

Versions tell the same story from another angle. We currently see 216 distinct version strings in production, with 2.4.68 the most common. A distribution that wide is the signature of software installed once and left alone, and it is precisely why the head of the distribution is the least useful part. Advisories apply to ranges, and the population sitting inside a vulnerable range is what a security team, an underwriter or an acquirer actually needs to size.

That is the practical limit of any summary statistic about this technology. Knowing the total tells you nothing you can act on. Knowing which domains in a specific country run a version below a specific threshold, behind a specific proxy, on a specific TLD, is a filter over a full population, and it either exists as a downloadable list or it does not exist at all.

The webatla export is that list. Every domain we detect it on, with the version string we observed, the hosting country, the TLD and the other technologies we fingerprinted on the same site, as a single file rebuilt daily. There is no per-query cost and no cap on how specific your filter gets, because the filtering happens on your machine rather than ours.

One consequence of that inertia is worth spelling out for anyone sizing a market. The population does not churn the way a product's user base churns. It erodes slowly at the edges as hosting estates are retired, and it is replenished by default images that nobody chose. So a campaign aimed at this population is not competing against a purchasing decision. It is competing against the absence of one, which is a different and generally slower sales motion.

The other thing the full list makes possible is negative filtering. Knowing which sites run this server is useful. Knowing which sites in a segment do not, and what they run instead, is how you size a migration opportunity or a competitive displacement. Both directions come out of the same file, and neither is expressible as a per-domain lookup.

Apache Software Foundation publishes it under the Apache 2.0 license and the .com zone carries the largest single slice. As with anything distributed freely, no vendor holds a customer list, so measurement is the only route to a population count. That is the gap this dataset is built to close.