A domain extension is supposed to be a label. In practice it predicts the infrastructure behind the site remarkably well — because the extension a business buys and the provider it hires are both chosen by where it is.
We resolved DNS for every origin in the Chrome UX Report that our corpus covers, and looked up which mail, DNS and hosting providers the records point at. The twelve largest extensions are below.
Who provides the infrastructure, by extension
Share of each extension's scanned origins whose DNS points at that provider. Read across a row for a country's habits; read down a column for a provider's reach.
| Cloudflare | Microsoft 365 | Google Workspace | GoDaddy | Amazon Route 53 | Hostinger | IONOS | Google Cloud DNS | |
|---|---|---|---|---|---|---|---|---|
| .com | 26.8 | 21.3 | 20.4 | 14.3 | 9.8 | 4.2 | 1.3 | 4.3 |
| .org | 27.4 | 29.3 | 24.0 | 15.8 | 5.0 | 2.5 | 1.4 | 2.5 |
| .de | 7.0 | 26.9 | 3.6 | 9.2 | 2.2 | 0.7 | 18.9 | 0.8 |
| .br | 30.1 | 17.2 | 19.1 | 1.6 | 10.3 | 11.3 | 0 | 1.7 |
| .jp | 4.1 | 16.0 | 14.0 | 0.2 | 26.7 | 0 | 0 | 0.9 |
| .net | 31.6 | 12.6 | 12.0 | 8.0 | 14.1 | 2.5 | 1.1 | 2.4 |
| .ru | 8.0 | 3.4 | 2.2 | 0 | 0.1 | — | 0 | 0.6 |
| .uk | 24.9 | 39.5 | 14.4 | 18.2 | 4.4 | 2.1 | 5.6 | 2.4 |
| .it | 15.5 | 21.4 | 15.3 | 1.4 | 3.0 | 1.1 | 0.9 | 1.1 |
| .fr | 10.2 | 22.7 | 6.8 | 0.6 | 2.3 | 3.2 | 7.2 | 1.7 |
| .pl | 23.4 | 21.6 | 14.9 | 0.6 | 4.3 | 1.3 | 0 | 0.7 |
| .in | 19.4 | 10.8 | 20.8 | 23.9 | 6.9 | 20.8 | 0.1 | 1.8 |
Source: DNS records for every origin in the Chrome UX Report that carries DNS evidence in our corpus. Percentages are of that extension’s scanned origins, not of its registrations.
The web is more national than it looks
.de is the clearest case. IONOS — a Montabaur company — serves 18.9% of German-extension sites and 1.3% of .com. In the other direction, Cloudflare reaches 26.8% of .com and only 7.0% of .de. Google Workspace, which is on a fifth of .com, is on 3.6% of .de. German sites buy German infrastructure.
.jp made a different choice and made it just as uniformly: Amazon Route 53 serves 26.7% of .jp origins, while Cloudflare manages 4.1%. .ru is the negative case — every provider in this study is in single digits there, because Russian sites use Russian providers that do not appear in a list of eight Western ones.
Cloudflare's reach, by extension
One provider, twelve extensions. The spread between the top and the bottom is nearly eightfold.
Source: DNS records for every origin in the Chrome UX Report that carries DNS evidence in our corpus. Percentages are of that extension’s scanned origins, not of its registrations.
Microsoft owns the mail, except where it does not
Microsoft 365 is the most widespread provider in the study, and .uk is its stronghold: 39.5% of British-extension sites route mail through it — nearly three times Google Workspace’s 14.4%. On .org the two are much closer, 29.3% against 24.0%.
.in inverts it. Google Workspace leads at 20.8%, Microsoft trails at 10.8%, and the real story of that row is elsewhere: GoDaddy at 23.9% and Hostinger at 20.8%. Nearly half of Indian-extension sites sit on budget shared hosting, a concentration no other extension comes close to.
Why this study is about DNS and nothing else
The obvious version of this post is “WordPress by extension”. We did not write that one, because our data cannot support it honestly, and the reason is worth stating.
- Evidence has sources, and they do not cover the same sites. DNS resolution runs against our whole corpus. The browser scan that detects a CMS has reached a much smaller, and not randomly chosen, subset.
- The gap is not small. Microsoft 365 has 4,909,552 detections from DNS and 1,568 from the browser scan. WordPress has 601,322 from the browser scan and none at all from DNS.
- So a mixed chart would measure the scanner. Divide both by a denominator built from DNS coverage and WordPress reports near zero everywhere — a chart that looks fine, renders fine, and says nothing true.
Restricting to one source costs this post its CMS column and buys every remaining column the right to be compared. A CMS study needs its own denominator, and it is a separate piece of work.