Redirects done properly: 301s, chains, and migrations

In my website migration checklist I said the redirect map is most of the job, and then moved on, because that post had four other phases to cover. This is the detail behind that line. Redirects are the plumbing of technical SEO, invisible when they work and expensive when they don't, and in audits they're one of the most common places I find rankings quietly leaking away years after anyone last thought about them.

The good news is that the rules are few, they're stable, and Google has been unusually clear about most of them. The bad news is that the mistakes are just as few and people make them anyway.

What a redirect actually tells Google

A redirect is a signal, not just a forwarding address. The status code you choose tells Google how confident to be that the move is real, and how quickly to transfer everything the old URL earned to the new one.

Type	What it says	When to use it
301 / 308 (server-side)	Permanent move, strong signal	The default for any URL that has genuinely moved
302 / 307 (server-side)	Temporary move, weaker signal	Genuinely temporary situations only, e.g. a page briefly redirected during maintenance
Meta refresh / JavaScript redirect	A hint at best	Last resort when you can't touch the server, and even then reluctantly

Google's documentation treats server-side permanent redirects as the strongest canonicalisation signal available, with temporary and client-side redirects sending progressively weaker ones. In practice Google often works out that a long-standing 302 is really permanent, but "Google will probably figure it out" is not a strategy. If the move is permanent, say so with a 301.

The PageRank question, settled

For years the industry believed redirects burned a slice of link equity on every hop, which drove some odd behaviour, including people avoiding redirects that plainly needed to exist. Google put this to bed in 2016, when Gary Illyes confirmed that 30x redirects don't lose PageRank, any of them.

So the equity question is settled, but it settled less than people think. A redirect passes its value to the destination only when the destination is a genuine equivalent. Redirect an old product page to your homepage, or to something loosely related, and Google is liable to treat it as a soft 404, at which point the equity doesn't transfer anywhere. The redirect type stopped being the risk. Where you point it never did.

Chains, where the waste actually happens

Redirect chains are what happens when nobody keeps the map tidy. HTTP redirects to HTTPS, non-www redirects to www, the old URL redirects to the 2023 structure, and the 2023 structure redirects to last year's rebuild. Four hops, each added by a different person solving a different problem, none of them wrong on its own.

Google's documentation says Googlebot follows up to ten hops before giving up, so a chain of three or four still resolves. The costs are quieter than a hard failure. Every hop adds latency for real users, chains burn crawl capacity on large sites, and a chain is fragile, because any one link breaking takes out every URL routed through it. The fix is one of the more satisfying jobs in technical SEO: crawl the site, find the chains, and repoint every redirect straight at the final destination. One hop, always.

The same applies inside the site. Internal links pointing at redirected URLs work, but they route every click and every crawl through an unnecessary hop, so when a URL changes, update the internal links to the new address rather than leaning on the redirect. Finding both problems is bread-and-butter work in a proper technical audit, and a crawler surfaces them in minutes.

How long to keep redirects

Longer than you think, and Google has been specific. John Mueller's guidance is to keep redirects in place for at least a year, because Google's systems need to see a redirect several times over a period of months before the move is fully processed. My own rule is stricter, because Google isn't the only thing that matters. Old URLs live on in bookmarks, in other sites' links, and in printed material, and every one of those is a visitor who hits a 404 the day the redirect rule gets dropped.

So treat redirect rules as permanent infrastructure, not temporary scaffolding. Rules from previous migrations should survive replatforming, which is exactly where they usually die, quietly dropped because the new platform's config started from a blank file. It's one of the first things I check when a migration has gone wrong and nobody can say why, and preserving them is part of how I plan the site builds I take on.

The mistakes I actually see

After enough audits the same short list keeps coming up. In rough priority order, the ones that cost real traffic are mass redirects to the homepage after a migration (soft 404s, equity gone), legacy redirect rules dropped during a replatform, chains left to accumulate across successive rebuilds, and 302s left in place for years on permanently moved URLs. None of them is hard to fix. All of them are hard to notice, because a redirect problem never announces itself, it just shows up as a slow bleed in Search Console that everyone attributes to something else.

That's the real argument for doing redirects properly. Not that any single rule is difficult, but that the cost of getting them wrong arrives months later, unlabelled, when the cause is hardest to see. If a site has been through a move and traffic never quite recovered, the redirect layer is where I look first, and it's core technical SEO work to put right.

301 redirect FAQs

Do 301 redirects lose PageRank?

No. Google confirmed in 2016 that no 30x redirect loses PageRank. The caveat is relevance, because a redirect to a page that isn't a genuine equivalent can be treated as a soft 404, and then nothing transfers. Point redirects at the closest matching page, not the homepage.

Should I use a 301 or a 302?

If the move is permanent, a 301. A 302 tells Google the original URL will return, so Google keeps treating it as the canonical one for longer. Google often reclassifies long-standing 302s as permanent, but there's no reason to make it guess. Reserve 302s for situations that are genuinely temporary.

How many redirects can I chain together?

Googlebot follows up to ten hops, but treat that as a limit, not an allowance. Every hop adds latency and fragility, so repoint chained redirects straight at the final destination whenever you find them. One hop is the standard to hold.

When can I remove old redirects?

Google's guidance is to keep them for at least a year, and I'd keep them indefinitely wherever the old URL has backlinks or bookmark traffic. Redirect rules cost almost nothing to keep. The traffic they protect can be significant, and the classic mistake is losing the rules accidentally in a replatform rather than removing them deliberately.