What we did
On 19 August 2026 we took the
Tranco top-1 million list
published 17 August 2026 (list ID K9L5W), filtered it to
every domain ending in .au — 8,376 of them — and
requested each homepage once over HTTPS, parsing the head with the same engine behind
the checker on our homepage. 6,445 returned a page we could measure,
and every percentage below is a share of those.
The rest are excluded rather than counted as failures of the sites themselves.
1,753 never returned a page —
633 blocked (http 403), 217 timed out, 241 tls problem, 352 not publicly resolvable, 85 other http error, 225 other.
A further 178 answered with HTTP 200 but served a bot challenge rather than a
real page: no title, no description, no images, no headings. We identify ourselves honestly
in the User-Agent instead of impersonating a browser, and some sites answer that with an
interstitial — amazon.com.au returns two kilobytes containing
<title> </title>. Counting those as "has no meta description"
would be false, and would bias the whole report against exactly the large sites most likely to
sit behind a firewall, so they are excluded and counted here instead.
Title tags
Search engines truncate titles at roughly 60 characters. Under 30 usually means the title is doing no work at all.
Meta descriptions
Google rewrites descriptions often enough that a missing one is not fatal — but it hands the decision entirely to the search engine.
The tags almost everyone gets right, and the ones almost nobody does
2,141 homepages (33.2%) have no Open Graph tags at all. Every time one of those is shared to Facebook, LinkedIn, Slack or iMessage, the platform guesses — usually badly — at a title, a description and an image. It is the cheapest fix on this page and the most visible when it is missing.
Headings
One H1 per page is the convention. Both zero and several are common enough to be worth counting rather than assuming.
Alt text
We counted 358,471 images across 6,141 homepages that carry any. 132,897 of them (37.1%) have no alt attribute. Only 1,077 homepages (17.5% of those with images) get every image labelled.
This is the one number here that is an accessibility problem before it is an SEO problem. A screen reader announces an unlabelled image by its filename, or skips it.
Grades
The grade each homepage would receive from the checker on our homepage, using the identical grading code.
What Australian websites are built with
Read from the generator meta tag. Only 948 of the
6,445 homepages we fetched (14.7%) name a platform we recognise —
most sites publish no generator tag at all, and some platforms deliberately omit it. So this is
a picture of the sites that announce themselves, not of the whole population, and the shares
below are of those 948 domains only.
Does the platform decide the outcome?
The same measurements, split by the platform a site declares. A CMS that ships good defaults should show up here.
| Platform | Domains | Description | Open Graph | Canonical | Grade A or B |
|---|---|---|---|---|---|
| WordPress | 271 | 62.0% | 36.9% | 91.1% | 22.9% |
| Elementor | 264 | 75.4% | 65.5% | 98.5% | 43.2% |
| Drupal | 250 | 52.0% | 15.2% | 98.4% | 13.6% |
| HubSpot | 37 | 97.3% | 54.1% | 100.0% | 64.9% |
| Webflow | 27 | 96.3% | 3.7% | 59.3% | 11.1% |
| Gatsby | 26 | 88.5% | 26.9% | 69.2% | 26.9% |
Platforms with fewer than 25 declaring domains are omitted.
By type of Australian domain
| Namespace | Domains | Description | Open Graph | Canonical | Viewport |
|---|---|---|---|---|---|
com.au |
4,870 | 83.6% | 44.8% | 79.2% | 95.4% |
au (direct) |
506 | 94.5% | 5.7% | 94.7% | 98.8% |
org.au |
353 | 72.8% | 43.1% | 77.6% | 97.7% |
gov.au |
304 | 59.5% | 26.0% | 75.3% | 96.4% |
edu.au |
201 | 78.1% | 41.8% | 86.6% | 99.0% |
net.au |
195 | 72.3% | 35.4% | 65.1% | 83.1% |
Methodology
- Population
- Every domain ending in
.auin the Tranco list published 17 August 2026 (IDK9L5W, permanently retrievable at https://tranco-list.eu/list/K9L5W) — 8,376 domains, the whole.ausubset rather than a sample. - Measurement
- One HTTPS GET of
https://<domain>/per domain, following up to five redirects, with a 15-second timeout and a 2 MB cap on the HTML parsed. We identify ourselves in the User-Agent. Only the homepage is measured, so this describes front doors, not whole sites — and a homepage is usually the most carefully tuned page a site has, which means these figures are likely the optimistic end of each site's practice. - Denominators
- Unless stated otherwise, percentages are a share of the 6,445 homepages we measured, not of the 8,376 in the list. Domains that failed to respond, and those that answered with a bot challenge, are excluded rather than counted as missing tags — counting a blocked request as a missing description would understate the whole population and would do it unevenly.
- The bot-challenge exclusion, and its risk
- A challenge page returning HTTP 200 is indistinguishable from a real page except by its contents, so we classify a response as a challenge when it has an empty title AND no description AND no Open Graph AND no images AND no H1 — five things a genuine homepage essentially never lacks at once. This is a heuristic, not a certainty. It will misclassify a truly empty holding page as "blocked", and it will miss a challenge page that happens to carry a title. We would rather exclude a handful of genuinely bare pages than publish an aggregate quietly dragged down by pages we were never shown.
- Alt text counts images, not decisions
- We count
imgelements with noaltattribute. An emptyalt=""is the correct, deliberate markup for a decorative image and is counted as present, not missing. CSS background images and inline SVG are not counted at all. - Why there is no structured data section
- The engine behind this report does not parse JSON-LD or microdata, so we have no honest measurement of schema.org adoption to publish. Rather than infer it from something adjacent, we have left it out. It is the obvious next thing to add.
- Platform detection is a floor
- Read from the
generatormeta tag only. We do not fingerprint scripts, paths or headers. Most sites publish no generator tag, and several platforms strip it, so the platform counts describe self-declaring sites and understate every platform — some far more than others. Do not read them as market share. - Point in time
- A single fetch per domain on 19 August 2026. A site mid-deploy, geo-serving different markup, or rendering its head in JavaScript after load may be measured differently than a browser would see it.
- 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 domain in the set can be re-checked by hand.
Corrections welcome. If a domain here is measured wrongly, or you can see a methodological problem we have missed, tell us and we will correct the page and say what changed.