Redirects that do not rot — one hop, or you have not moved the URL

A redirect is a promise to two parties: the person with a bookmark or a link in an email, and the crawler that holds a signal you would like to keep. The promise holds when the old hop lands on the right page in one 301. It rots when you stack a hop on a hop, leave a 302 standing, or answer 200 on something that is no longer a page. You think you have moved the URL. You have moved the pain to a place you do not look.

Chains are not a cosmetic problem. Each extra hop is crawl budget you donate, and a place a client or a bot can stop. An A→B→C where C later became 404 is three chances to lose what you thought you had saved. A loop is worse: you have asked the browser to stay. Teams notice when PageSpeed or a crawl tool shows 'redirect chain' on a URL they promised was canonical. They fix that one. They do not touch the inventory.

302 vs 301 is the choice you make when you are unsure, and forget to come back. A 302 says temporary. The search engine is allowed to treat it as that. If the job was 'the old slug is dead, the new one is home', 301 is the word. If the job was 'we are A/B-testing a landing page for two weeks', 302 is right — and you owe a date when it either becomes 301 or goes away. A 302 without an owner is how you have a temporary hop for three years.

We write this because 'we have redirects' has become a checkbox you tick at go-live and never open again. An honest setup is boring: one hop per old URL, 301 when the move is real, 404 when the page is dead, and a queue you actually work. Without that habit you have a file from the agency. You do not have an inventory.

A redirect is a named promise: this URL is no longer here; it lives there. The status code is part of the promise. 301 and 308 say permanent. 302 and 307 say temporary. The difference between 301/302 and 308/307 is whether the method may change — relevant for forms and API calls, rarely for an old blog post. For a marketing site the sentence is almost always: 301, one hop, to the canonical. Everything else is an exception you should be able to explain.

It is not a canonical tag. Canonical says 'of the pages that exist, this is the preferred one'. Redirect says 'this URL no longer exists as a page'. Mix the two and you have a URL that 200s with a canonical to another host while you also 301 a third place. The crawler chooses. You have not. Canonical is for duplicates you still serve. Redirect is for URLs you no longer want inside.

It is not a sitemap entry either. The sitemap tells what you want crawled now. The redirect inventory tells what you owe the URLs you no longer want crawled as themselves. Deleting a slug from the sitemap and letting it 404 without deciding that is not hygiene. 301ing every unknown path to the homepage is not hygiene either: it is a soft 404 with a 200 on the destination, and the crawler learns that your homepage is the answer to everything dead.

And it is not 'a rule in the WordPress plugin you toggle on'. A plugin that auto-301s on a slug change is a tool. It is not ownership. The plugin remembers the last change. It does not remember last autumn's campaign, the PDF you moved, or that you now run Next.js and the old .htaccess is no longer read. The inventory is the file you deploy. Everything else is a hope you parked with a vendor.

The first failure is stacking hops. You change the slug again. The plugin or next.config adds B→C. A→B stays. A year later it is A→B→C→D, and D is a 404 because someone cleaned 'unused pages'. The relaunch post is having the map on cutover day. Here the duty is to collapse the chain every time you move. An inventory you only open at relaunch is how redirects rot.

The second failure is the 302 you never made permanent. 'We are not sure about the new URL.' That is honest for two weeks. It is a lie for two years. The search engine was told the move is temporary. You were told you would come back. Nobody did. Write the date, or write 301. A 302 without an owner is a temporary hop you have turned into architecture.

The third failure is regex that swallows what you did not mean. A catch-all that drops the query string, so UTM and pagination land wrong. A trailing-slash rule that chains with the www rule that chains with http→https, so one bookmark is four hops. Write the few you owe, explicitly. A clever star is how you 301 `/blog/foo` to `/` because you meant `/blog/:slug`. Test with the URLs you actually have out in emails.

The fourth failure is the soft 404 and the homepage funnel. An SPA that 200s every path and renders 'page not found' in the client. A WordPress theme that shows a search result with 200. A rule that sends unknown URLs home 'so we do not lose the link'. You have told the crawler it is a page. It is not. The error-page post is chrome. Here the duty is to let a dead URL be dead — 404 or 410 — or to move it in one 301 to something that actually replaces it.

Relaunch is the job that is most often right as a 1:1 map, and most often rots as a spreadsheet. The relaunch post is parity and cutover. Here the duty afterwards: every new slug change collapses the chain. You are not moving 'the site'. You are moving the URLs you still owe. If the new stack does not read the old .htaccess, cutover is not done. You have changed host. You have not moved the hops.

Campaign URLs are the job you forget because they do not look like 'the site'. /spring-offer, /lp/webinar-may, a UTM landing you promised in a newsletter. When the campaign is dead, the URL is too — unless you 301 to what actually replaces it, or 404 on purpose. Leaving it 200 as an empty template is a soft 404. Sending it home is telling the crawler the homepage is the offer. Write the replacement, or write 404.

www, https and trailing slash are the job that most often becomes four hops. http://www.example.com/foo/ → https://www… → https://example.com/foo/ → https://example.com/foo. That is one canonical URL you have turned into a chain. Set the canonical. Let the other three 301 straight there. The performance post is LCP. Here is refusing to let a bookmark pay four RTTs for a slash you would not own as content.

Old PDFs, attachments and 'download the price list' are the job that most often 404s in a mail someone forwards again. If the file moved, it is a 301 to the new file or to the page that replaces it. If the file is dead on purpose, it is 404 — not a hop to the homepage so you can say the link 'works'. A document you have shared in the world is a URL you owe an answer. It is not 'just a file in /public'.