Redirecting old URLs after a relaunch: what worked and what broke

What I learned setting up redirects for metahead's relaunch: one pattern rule, a cache trap, 410s, redirect chains, and why Google may not follow a redirect right away.

Published Nov 16, 2026 · 7 min read

When I relaunched metahead this month, the hardest part wasn’t the new design. It was the old URLs. The blog used date-based addresses like /blog/2010/03/20/post-name/, and the new site uses short ones like /post-name/. Every link other sites had pointed at an old post was aimed at an address that no longer existed.

Redirects fix that, and the basic idea is simple: a 301 tells browsers and search engines that a page has moved for good. What the guides leave out is how messy it gets in practice. Here’s what I did on metahead, what broke, and what I’d do again.

What a 301 actually tells Google

A 301 is a permanent redirect. Google’s documentation says it treats a permanent redirect as a signal that the new address should be the one it shows in search results, while a temporary redirect, like a 302 or 307, keeps the old address showing. A 308 counts as permanent, too.

Google also says a server-side redirect has the best chance of being read correctly, so set it up on the server or in a plugin that answers at the server level, not with JavaScript.

Notice the word signal. A redirect isn’t an order, and that matters later in this post.

Step 1: Find the old URLs

You can’t redirect what you don’t know about. Google’s site-move guide suggests checking your server logs for old URLs that people still visit.

My logs from the 2000s were long gone, so I used Semrush’s list of metahead pages that other sites link to. It turned up about 400 old URLs: roughly 115 posts and monthly archives, and the rest old tag pages.

That list told me what was worth saving. The posts were the pages with real content behind those links. The tag pages were a different case, which I’ll come back to.

Step 2: One pattern rule instead of 100 redirects

Because the old and new addresses share the same slug, I didn’t need a separate redirect for every post. One pattern rule, a regex (short for regular expression), in Rank Math handles all of them:

Source (Regex): ^blog/[0-9]{4}/[0-9]{2}/[0-9]{2}/(.+)$
Destination: /$1
Type: 301

It takes anything shaped like /blog/year/month/day/post-name/ and sends it to /post-name/ with a 301. In under four days it was used 161 times. That count includes bots and crawlers, but every one of those requests would otherwise have hit a 404.

One Rank Math quirk: Rank Math’s support team says its regex redirects don’t support a forward slash at the start of the source, so for an anchored rule like this one, write ^blog/…, not ^/blog/…. Some unanchored examples in Rank Math’s regex guide do start with a slash.

Step 3: Test it as a visitor, and clear your cache

When I tested the old URLs after adding the rule, some still came back as “not found.” The rule was fine. My caching plugin was serving saved copies of those 404 pages from before the redirect existed. After I cleared the cache and tested again as a logged-out visitor, they redirected properly.

If you use a page cache, purge it after adding redirects, and test logged out, the way search engines see your site.

Step 4: Decide what doesn’t come back

Not every old post deserved a new home. I dropped a dozen, mostly expired coupon codes and deal posts, plus a couple of posts about a boxing match, an old Super Bowl ad post, and a Windows 7 how-to.

There’s no good page to send those visitors to, and sending them all to the homepage wouldn’t help anyone looking for a coupon from 2010.

Those return a 410, which means “gone for good.” In practice, Google treats a 410 like a 404: its documentation says all 4xx errors except 429 are treated the same, and URLs that return one are dropped from the index.

I use 410 anyway because it records a decision. A 404 in my logs means something I haven’t dealt with yet. A 410 means I did. Those 12 URLs have been requested 259 times since I added the rule on November 9.

The old tag and monthly archive pages return 404 now, too. They were lists of links, and no single page replaces them.

Step 5: Let WordPress handle slug changes

A few minutes after I published my build stack post, I changed its address from /build-stack/ to /wordpress-build-stack/. I didn’t add a redirect for it, because WordPress already had. When you change a published post’s slug, WordPress looks up the old slug and sends visitors to the new address with a 301.

That’s worth knowing before you add a plugin redirect on top. Two rules for the same URL is how chains start.

Step 6: Watch for chains

A chain is a redirect that lands on another redirect. Each hop costs time. Chrome’s Lighthouse documentation explains that after a redirect, the browser has to make another request at the new address, and that extra trip can delay loading by hundreds of milliseconds. A chain makes visitors wait for it more than once.

That matters because people leave slow pages. Google’s web.dev cites the BBC losing an extra 10% of users for every additional second its site took to load.

Google says it uses Core Web Vitals, which include how fast a page loads, in its ranking systems, but it still shows the most relevant content first, even when the page experience is sub-par. One extra hop is unlikely to sink you. A pile of them is still an easy fix that makes the site faster for everyone who clicks an old link.

Google’s crawlers follow up to 10 redirect hops, so a short chain won’t break anything, but Google’s site-move guide advises redirecting straight to the final page, and keeping any chain to ideally no more than three hops.

I made one this week without noticing. I merged two old posts about PayPal restrictions in India and redirected the retired post to the one I kept. An old shortlink, /go/paypal, still pointed at the retired post, so it now went through two redirects. The fix was to point the shortlink straight at the final post.

Some chains aren’t worth fixing. Old dated links to that retired PayPal post now take two steps as well: the pattern rule sends them to the old slug, and the merge redirect sends them on.

On my site, Rank Math runs the pattern rule first, and carving exceptions into it would be fragile. Two hops is well inside Google’s limit, so I left it.

Step 7: Merging posts is a redirect decision, too

The relaunch brought back posts I wrote between 2008 and 2014, and two pairs covered the same ground: a 2008 spam email and my follow-up about it, and two 2011 posts on PayPal’s restrictions in India. Two thin pages on one topic split attention, so I merged each pair.

The better post kept its URL and its original date and took the useful part of the other, and the other got a 301 to it.

I set the retired posts to draft rather than deleting them, so the text is still there if I need it. I also fixed the one internal link that pointed at a retired post, because Google’s guide says to change internal links to the new URLs instead of relying on the redirect.

When Google ignores your redirect

Here’s the part I’m still waiting on. The build stack redirect works, but when I checked Search Console on November 12, Google had marked my new URL as a duplicate, “Google chose different canonical than user,” and listed the old /build-stack/ address as the page that referred it. Nothing had changed when I checked again on November 13.

That’s the “signal, not an order” point in action. Google calls a redirect a strong signal for which URL should be the main one, but the final call is still Google’s.

I’ve confirmed the redirect is permanent, made sure no internal links point to the old address, and asked Google to crawl the new one again. Now I wait, and I’ll update this post when it switches.

How long to keep redirects

Longer than you think. Google’s site-move guide says to keep redirects for as long as possible, generally at least a year, and suggests keeping them indefinitely for your visitors’ sake. Old links still get used, too: the pattern rule alone had 161 hits in under four days, so mine stay.

My redirect checklist for a relaunch

Here’s the short version, in the order I’d do it again:

  • List your old URLs from server logs or a backlink tool.
  • Use one pattern rule when old and new addresses share a slug, and one-to-one redirects for the rest.
  • Use a 301 for pages that moved, and a 410 for pages you’ve retired on purpose.
  • Purge your cache and test every old URL as a logged-out visitor.
  • Don’t stack a plugin redirect on top of WordPress’s own slug redirect.
  • Point redirects at the final page, and recheck old shortlinks after merging posts.
  • Update internal links to the new URLs.
  • Keep redirects for at least a year, and longer if old links still get used.
  • Check Search Console, and give Google time.

Setting up the redirects was the easy part. Checking what they actually did is where I found every problem in this post.

Top photo: Carlos Macías (CC0).

Leave a comment