Corrected 2 September 2026.
An earlier version of this report overstated how many links go through a redirect. Our
scanner followed each homepage's redirects but then resolved that page's relative links
against the address we had asked for rather than the one we landed on — so on any site that
sends example.com to www.example.com, every internal link was
recorded as a redirect it never actually had, and counted as external rather than internal.
The previous figure of 54.5% of links going through a redirect was wrong for that reason. The corrected figure, from a fresh scan, is 10.9%. Broken-link counts were affected far less, because a mis-resolved link still resolved to a real page on a real host, but every number on this page comes from the new scan. The bug is fixed in the checker on our homepage too, where it had the same effect on individual scans.
What we did, and why this one is a sample
On 2 September 2026 we took the most popular .au domains from the
Tranco top-1 million list
published 1 September 2026 (list ID K9QPW), fetched each
homepage, extracted its links and requested up to
15 of them, using the same engine behind
the checker on our homepage.
This report is deliberately sampled, and the other reports in this suite are not. Checking a page's links means sending requests to whoever that page links to — third parties who never asked to be crawled. Running our full list of 8,376 Australian domains at the tool's normal 150 links per page would be well over a million outbound requests made in our name. So we took the top of the list and capped links per page hard. 710 homepages, 10,045 links. Read this as a study of Australia's most-visited sites, not of the whole namespace — and expect the true figure across the long tail to be worse, since these are the sites with the most resources to maintain themselves.
Of 1,000 domains sampled, 710 gave us a page with links to check. 252 could not be fetched at all, and 38 returned a page with no links — usually a bot challenge served to our honest User-Agent rather than a genuinely linkless homepage. Both are excluded rather than counted as clean.
What happened to the links
A redirect is not a fault — but at 10.9% of all links, it is a lot of avoidable round trips, and every one is a link that could be pointed at its final destination instead.
How many sites are affected
Grades
The grade each homepage would receive from the checker on our homepage, using the identical grading code.
What Australian sites link to that has gone away
The destinations that came back broken most often, counted by how many different Australian homepages still link to them. Hosts only — we are not publishing a list of specific dead pages on other people's sites.
By type of Australian domain
| Namespace | Homepages | Links checked | Broken links | Sites with a dead link |
|---|---|---|---|---|
com.au |
480 | 6,736 | 7.4% | 26.5% |
gov.au |
94 | 1,407 | 3.7% | 14.9% |
edu.au |
56 | 829 | 11.1% | 28.6% |
org.au |
43 | 636 | 4.1% | 23.3% |
net.au |
28 | 364 | 6.3% | 25.0% |
Namespaces with fewer than 20 measured homepages are omitted.
Methodology
- Population and sampling
- The most popular
.audomains in the Tranco list published 1 September 2026 (IDK9QPW, permanently retrievable at https://tranco-list.eu/list/K9QPW). Unlike the other reports in this suite this is a sample, not the full namespace, for the reason given above: the outbound request cost falls on third parties. - Measurement
- Homepage fetched, links extracted from the HTML, and up to 15 of them requested. A link counts as broken on a 404, 410 or similar not-found response, and as a timeout when nothing came back inside the engine's limit. Redirects are followed and reported as redirects rather than failures.
- Only the homepage, only the first links
- We check the homepage of each site and the first 15 links on it, in document order. That biases towards navigation and header links, which are the most maintained links on any site. Links buried in old articles — where rot actually concentrates — are not represented here at all. The real rate across a whole site will be higher than this page reports.
- A timeout is not proof of death
- A destination that did not respond inside our limit may be slow, rate-limiting us, or blocking automated requests rather than gone. We report timeouts separately from broken links for exactly that reason, and they should not be added together without saying so.
- Point in time
- A single check per link on 2 September 2026.
- Reproducing it
- The aggregated results are published as JSON under a CC BY 4.0 licence. The Tranco list ID pins the exact input, and the scanning code is the same engine behind the homepage checker, so any page in the set can be re-checked by hand.
Corrections welcome. If a link here is judged wrongly, or you can see a methodological problem we have missed, tell us and we will correct the page and say what changed.
Related reports
Because this study samples the top of the list, it cannot speak for the whole namespace. Two companion reports go deeper on narrower questions: a full census of Australian government, university and charity sites, and an audit of what all the redirects we keep finding are actually for.