Registry vs reality: one address in eight points at nothing
DNS check, taken 2026-09-01 07:03. Data and tooling are open.
What was measured
Of the 11 902 domains appearing in our crawl log of US municipal websites, 1 498 (12.6 %) have no DNS record. The name does not exist.
This is not "the site failed to load" or "the server is down". It means the domain was not renewed and stopped existing as an address. Typing it into a browser is like dialling a phone number that was never assigned.
All of them are still listed as working addresses in the federal registry the sample was drawn from.
Numbers
How this differs from "the site did not respond"
The data page reports that some addresses did not respond to the crawler. That is a fact about server behaviour at one moment: it may have been overloaded, timed out, or dropped the connection. An hour later the same address might have worked.
Here we check whether the name itself exists, not whether a server answers. Different questions, different numbers — and they must not be mixed.
| Question | What a "no" means |
|---|---|
| Did the site load? | The server did not answer now. It may answer later. Fixed by configuration. |
| Does the domain exist? | The name is absent from the DNS. It will not start working later. Fixed only by registering it again. |
How it was checked
For each domain, A and AAAA records were requested over DNS-over-HTTPS. A domain counts as dead only when both resolvers return nothing.
Two independent resolvers were used — Cloudflare and Google. A single resolver can be wrong, be blocked in the country of testing, or serve a stale cache.
The disagreement between the two resolvers is the error margin of the method. It is measured and stated below, not estimated by eye.
DNS queries are free and place no load on the sites under study: not a single request was sent to the servers themselves.
What follows from this
First: the share of accessible sites depends on what you take as the denominator. Counted against every address in the registry, fewer journeys reach the goal than when counted against sites that actually exist. We publish both numbers and state the difference — otherwise the result would either flatter the sites or overstate the barrier.
Second, and more important than accessibility: the federal registry is not kept current. One address in eight leads nowhere, and nobody noticed.
Third, and this one is troubling: an expired domain can be registered by anyone. An address people take for governmental, listed in a federal registry, may one day come back to life — showing anything at all.
Limitations
- A missing A/AAAA record does not always mean abandonment: a domain may serve mail only (MX records) or be used internally. Such cases are possible in our sample and their share was not measured.
- The check is a snapshot of one moment. A domain could have been temporarily undelegated and later restored.
- The sample consists of domains appearing in our crawl log, not the entire federal registry.
- We did not check who currently owns the expired domains. The re-registration risk is stated as a possibility, not as observed cases.
How to verify this yourself
Any domain on the list can be checked with a single terminal command or through any public DNS service. The result reproduces independently of us.
The tool that produced these numbers is open and runs with no keys and no payment.
Raw data
The complete list of 1,498 domains as CSV: domain name and how many crawl records belong to each. Opens in Excel with a double click. The date in the filename is deliberate: a DNS check is a snapshot, and a domain may come back.
Author and disclosure
Maksim Valentinovich Galatin.
Conflict of interest disclosure: the author provides website accessibility and security auditing services. This research is self-funded; nobody paid to be included in the sample or to influence the result.