Redirect check

Why one page, one address

Every page should have exactly one address. Not because Google would penalise you, but because a split website is weaker. Links, history and rankings get divided between variants that are one and the same page to a human. When the homepage lives at four addresses, each one has a quarter of the strength.

Variants come about in four ways. firma.cz and www.firma.cz are two websites to a search engine. http:// and https:// are another two. An address with and without a trailing slash is the third pair, if both return content. And a non-existent address that returns 200 with the text "page not found" instead of a 404 is the fourth source of duplicates, because Google will index such a page.

What the tool measures

From the address you enter it derives the variants with http and https, with and without www, for a path it adds the variant with the opposite trailing slash, and finally it tries a made-up address on the same domain. It follows every variant hop by hop, without automatic redirect following, so you see the whole chain the way a search engine sees it. Every hop shows the status code and the Location header.

The verdict follows a simple rule. What is fine is a single 301 hop to one main version. Everything else is a finding.

• Two variants return 200. The website has two copies and the strength is divided.
• Code 302. It means temporary and Google keeps the original address. A permanent redirect has code 301 (or 308).
• A chain of two or more hops. http://firma.cz to https://firma.cz and from there to https://www.firma.cz/. Every step is another request and another delay; the ideal is a single hop to the destination. A common cause with http://www is the hosting provider, whose proxy switches http to https before the website sorts out www. The tool notes it next to the row, the fix is below.
• A non-existent page returns 200, or redirects to the homepage. A search engine reads both as a soft 404.
• A variant does not respond. Timeout, refused connection or an invalid certificate. For the http variant this is the same finding as a missing redirect.

What to do with the result

One button copies the table as plain text with the measurement date. It belongs in the audit report as evidence, not as an assessment. The brief for a supplier reads like this. "Redirect all variants (http, without www, with a trailing slash) with a 301 in a single hop to the main address. Every page should have a canonical address pointing to itself. A non-existent address returns 404."

One place, two rules

Redirects should live in one place, otherwise the hops add up. Most hosting providers offer a "redirect to https" switch that runs on their proxy before the request reaches the website, and it leaves www alone. The website then drops www only in a second hop. It is cleaner to turn the switch off and enforce both in the website .htaccess, the www rule first so that it points straight to https without www:

RewriteCond %{HTTP_HOST} ^www\.(.+)$ [NC] RewriteRule ^ https://%1%{REQUEST_URI} [R=301,L] RewriteCond %{HTTPS} !on RewriteCond %{HTTP:X-Forwarded-Proto} !https [NC] RewriteRule .? https://%{HTTP_HOST}%{REQUEST_URI} [R=301,L] 

The order matters. Deploy the rules first, only then turn off the switch at the hosting provider. The other way round the website is briefly reachable over unencrypted http with code 200, which is exactly what the tool reports as a duplicate for the http:// variant. The second condition on X-Forwarded-Proto is for hosting whose proxy terminates encryption and passes the request to the website unencrypted; without it the redirect would loop.

Where the tool comes from

It was born while writing the lesson One page, one address in the Technical website audit for beginners course in the ZSF club. The lesson explains why duplicate addresses weaken a website, how to find them in DevTools and how to brief a supplier on the fix. The tool is a shortcut to the same measurement from a single address.

The tool only checks public domains on standard ports. It will not go to an internal network, IP addresses or localhost, and it allows twenty checks per address every fifteen minutes. The probe sends a HEAD request with its own identifier and does not download the page body.