Keep the redirects at least a year. That is not a convention, it is the published guidance, and it is the single figure most migration checklists either omit or shorten.

The rest of the documentation is similarly specific: a named chain limit, an explicit statement that permanent redirects do not cost ranking credit, and a precise list of the moves for which the Search Console change of address tool is and is not required. That last list is where most published checklists are simply wrong.

This page gives each documented rule with its figure, and what to do when your case is not a URL change at all.

The redirect rules, with their numbers

Four items, each quotable.

Which redirect to use, verbatim. Use server side permanent redirects if technically possible. Although the crawler supports several kinds, the recommendation is HTTP permanent redirects if possible, such as 301 and 308.

Whether you lose credit, verbatim. Do not worry about link credit. 301 and other permanent redirects do not cause a loss in PageRank.

How long to keep them, verbatim. Keep the redirects for as long as possible, generally at least 1 year. That timeframe allows Google to transfer all signals to the new URLs, including recrawling and reassigning links on other sites that point to your old URLs. From users’ perspective, consider keeping redirects indefinitely.

The chain limit, verbatim. Avoid chaining redirects. The crawler can follow up to 10 hops in a chain, but the advice is to redirect to the final destination directly. If that is not possible, keep the number low, ideally no more than 3 and fewer than 5.

Why the retention figure matters most. Redirects get removed for mundane reasons: a hosting renewal, a server cleanup, a new team inheriting a configuration nobody documented. Doing that at nine months is inside the window Google names as necessary for signals to transfer.

The practical instruction that follows. Put a dated note in whatever system your team actually reads, saying which redirects exist, why, and the earliest date they may be reconsidered. Not a wiki page nobody opens.

What the tool covers, and what it does not

This is the section most migration guides get wrong, and the documentation is unambiguous.

When you need it, verbatim. If you are changing domain names or subdomains, submit a change of address in Search Console for the old site. You only need this tool when moving from one domain or subdomain to another, such as example.com to example.net, or a.example.com to b.example.com.

When you do not need it, verbatim. You do not need it for HTTP to HTTPS moves, for switching between www and non-www on the same domain, or for moving paths within the same domain.

Note what is on the covered list. Subdomain to subdomain. A common claim in migration guides is that the tool does not handle subdomains. It does, and the documentation gives it as an example.

What that leaves uncovered. Protocol changes, www normalisation, and path restructuring, all of which are handled by redirects alone.

Why the distinction is practical. Teams sometimes delay a protocol migration waiting for a tool step that does not exist for it, or expect the tool to smooth a path restructure it has no bearing on.

And what the tool never does. Substitute for the redirects. It is a notification, not a mechanism.

Published platform figures governing website migrationTable of the figures published by the search platform governing website migration. On redirect type, the platform recommends server side permanent redirects where technically possible, naming the three hundred and one and three hundred and eight status codes specifically, and states that three hundred and one and other permanent redirects do not cause a loss in ranking credit. On retention, it states that redirects should be kept for as long as possible and generally at least one year, because that period allows all signals to transfer to the new addresses including recrawling and the reassignment of links on other websites pointing to the old addresses, adding that from the user’s perspective redirects should be considered for indefinite retention. On chaining, it notes that the crawler can follow up to ten hops in a chain of multiple redirects but advises redirecting directly to the final destination, and where that is not technically possible keeping the number of redirects in the chain low, ideally no more than three and fewer than five. On timing, it states that for medium-sized websites it can take a few weeks or more for the platform to gradually begin showing new addresses instead of old ones, and even longer for larger sites, providing no upper bound. On the change of address tool, it states the tool is required only when moving from one domain or subdomain to another, giving both a domain-to-domain and a subdomain-to-subdomain example, and states explicitly that it is not needed for moves from insecure to secure protocol, for switching between the www and non-www forms of the same domain, or for moving paths within the same domain. On sequencing, it recommends that small and medium sites move all addresses simultaneously to help its algorithms detect the move and update the index faster, while large sites may move one section at a time to make problems easier to monitor, detect and fix.Every published figure, in one placeQuestionDocumented answerWhich redirect?Server-side permanent: 301 or 308Do I lose link credit?“301 and other permanent redirects don’t cause a loss in PageRank”How long do I keep them?At least 1 year. Consider indefinitely.How many hops can I chain?Up to 10 supported. Ideally ≤ 3, fewer than 5.How long until it settles?“a few weeks or more” for medium sites. No upper bound given.All at once, or in stages?Small and medium: all at once. Large: section by section.Change of address tool: needed forDomain to domain.Subdomain to subdomain. Yes, it covers those.Not needed forHTTP to HTTPS. www to non-www.Moving paths within the same domain.
At least a year on redirects, no more than three hops, and a tool that covers domains and subdomains only. Source : Google Search Central, site move with URL changes (2026)

Sequencing, and the fluctuation nobody warns clients about

Two more documented points, both of which change how you brief stakeholders.

On sequencing, verbatim. Small or medium sites: move all URLs simultaneously instead of one section at a time, because this helps users interact with the site better in its new form and helps the algorithms detect the site move and update the index faster.

For large sites. You can move one section at a time, which makes it easier to monitor, detect and fix problems faster. And separately: if your site is large and it is technically possible, consider moving just a piece first to test the effects on traffic and indexing.

On fluctuation, verbatim. Expect temporary fluctuation in site ranking during the move. As a general rule, for medium-sized websites it can take a few weeks or more for Google to gradually start showing the new URLs instead of the old ones, and for larger sites, even longer.

And the reassurance, verbatim. That visibility may fluctuate temporarily during the move, that this is normal, and that a site’s rankings will settle down over time.

Why this belongs in the brief, not the post-mortem. A traffic dip in week two of a migration is documented, expected behaviour. If nobody told the business that in advance, it becomes an emergency, and emergencies produce the worst decision available: reverting the migration halfway.

The number to commit to instead. Tell stakeholders that the comparison point is the same month next quarter, not next week. Then hold to it.

The documented post-migration fluctuation period and how to brief it in advanceDiagram presenting the post-migration ranking fluctuation period documented by the search platform and explaining why it should be briefed to stakeholders before a migration rather than explained afterwards. The documentation states that temporary fluctuation in site ranking should be expected during a move, that with any significant change to a site ranking fluctuations may occur while the platform recrawls and reindexes, and that as a general rule for medium-sized websites it can take a few weeks or more for the platform to gradually start showing the new addresses instead of the old ones, with larger sites taking even longer. It adds that the visibility of content in search may fluctuate temporarily during the move, that this is normal, and that a site’s rankings will settle down over time. No upper bound is published for large sites. The practical consequence is that a traffic decline during the second week following a migration constitutes documented and expected behaviour, but where the business has not been told this in advance it is perceived as an emergency, and emergencies during migrations produce the single worst available decision, namely reverting the migration partway through, which leaves the site in a state neither configuration anticipated. The recommended approach is to commit stakeholders in advance to a comparison point of the equivalent period in the following quarter rather than the following week, and then to hold to that commitment when interim figures look poor.Brief the dip before it happensWhat the documentation says, verbatim”for medium-sized websites, it can take a few weeks or more for Google to gradually start showingthe new URLs instead of the old ones (and for larger sites, even longer)“Briefed in advance”A dip is expected. We comparenext quarter.”Week two arrivesTraffic is down. Exactly asdescribed.Nothing happensWhich is the correctresponse.Unbriefed, the same week produces the worst decision availableReverting a migration partway, which leaves the site in a state neither configuration anticipated.Commit to the comparison point in advance: the same period next quarter. Then hold to it.
A dip in week two is documented, expected behaviour. Unbriefed, it becomes an emergency and produces a half-revert. Source : Google Search Central, site move guidance (2026)

When your move is not a URL change

A separate case with separate guidance, frequently conflated with the first.

What it covers. Switching hosting providers, or moving to a content delivery network. Guidance that explicitly applies only to migrations that do not affect the user-visible URL.

The steps. Prepare the new infrastructure, change the DNS, monitor traffic, then decommission the old infrastructure.

Which step is the move. The documentation identifies the DNS change as the actual site move step.

What is absent from this guidance. Redirects. There are none, because the addresses do not change.

Why people get this wrong. They apply the URL-change checklist to a hosting change, which produces a lot of unnecessary anxiety and occasionally a set of redirects pointing a URL at itself.

One thing that does still matter here. Server response time changes with hosting, and that is a real performance variable even when nothing about your addresses moved.

The two distinct categories of website move and their separate documented guidanceComparison of the two distinct categories of website move recognised by the search platform’s documentation, which are frequently conflated. The first category is a move involving address changes, covering situations where the user-visible addresses change, for which the documented steps are to prepare and thoroughly test the new site, prepare a mapping from current addresses to their new equivalents, begin the move by configuring the server to redirect from old addresses to new ones, and monitor traffic on both sets of addresses; server-side permanent redirects are required, retention is at least one year, chains should ideally not exceed three hops, and the change of address tool applies where the move is between domains or subdomains. The second category is a move without address changes, which the documentation describes as applying to a change in hosting infrastructure meaning switching hosting providers or moving to a content delivery network, and which it states applies only to migrations that do not affect the user-visible address. The documented steps for that second category are to prepare the new infrastructure, change the domain name system records, monitor traffic and then decommission the old infrastructure, with the documentation identifying the domain name system change as the actual site move step. Redirects are entirely absent from that second category because the addresses do not change, and applying the address-change checklist to a hosting change produces unnecessary anxiety and occasionally a configuration in which an address is redirected to itself. One variable does remain relevant in the second case, namely that server response time changes with hosting, which constitutes a genuine performance consideration even where no address has moved.Two kinds of move, two sets of rulesAddresses changePrepare and test the new siteMap old addresses to newConfigure the redirectsMonitor both sets of addresses301 or 308. At least a year. Max 3 hops.Change of address tool if domain or subdomain.Addresses do not changeNew host, or moving to a CDNPrepare the new infrastructureChange the DNSMonitor, then decommission the oldThe DNS change is the move.No redirects. There is nothing to redirect.Applying the first checklist to the second caseProduces a lot of anxiety, and occasionally a redirect pointing a URL at itself.One thing that does still matter on a host change: server response time is a real performance variable.
If your addresses do not change, there are no redirects and the DNS change is the move. Source : Google Search Central, site move guidance (2026)

The five failures worth planning against

Not from documentation, from what actually goes wrong.

Removing the redirects early. At a hosting renewal, a cleanup, or a handover. Inside the one-year window, this undoes the migration.

Chaining onto an earlier migration. Old URL to previous URL to current URL. Two migrations produce a two-hop chain, three produce three. The documented ceiling arrives faster than teams expect.

Redirecting everything to the homepage. Convenient, and it destroys the mapping. A page that no longer exists should return a 404 or 410, or redirect to its genuine equivalent, not to the front door.

Losing the old URL inventory. Once the old site is gone you cannot enumerate what it had. Export the full URL list and the Search Console performance data before you switch anything.

Forgetting non-page assets. PDFs, images referenced from other sites, feed URLs, and any address someone else hard-coded into an integration. These carry links and they break silently.

And the one governance failure behind most of them. Nobody owns the redirect map after launch week. Assign it, with the earliest review date being at least a year out.

Five common website migration failures and the point at which each occursTable of five common website migration failures together with the point in time at which each typically occurs and the governance failure underlying most of them. The first failure is removing the redirects early, occurring at a hosting renewal, a server cleanup or a team handover, which falls inside the one-year retention period the documentation specifies and consequently undoes the migration by preventing signals from completing their transfer. The second failure is chaining redirects onto an earlier migration, producing sequences from an original address to a previous address to a current address, where two migrations generate a two-hop chain and three generate a three-hop chain, meaning the documented ceiling of ideally no more than three hops arrives sooner than teams anticipate. The third failure is redirecting all removed pages to the homepage, which is convenient but destroys the mapping, whereas a page that no longer exists should return a four hundred and four or four hundred and ten status code, or redirect to its genuine equivalent, rather than to the front page. The fourth failure is losing the old address inventory, since once the old site is decommissioned its addresses cannot be enumerated, making it essential to export the complete address list and the search console performance data before switching anything. The fifth failure is forgetting non-page assets including portable document files, images referenced from other websites, feed addresses and any address hard-coded into a third-party integration, all of which carry links and break silently. The governance failure underlying most of these is that nobody owns the redirect map after the launch week, so the remedy is to assign ownership explicitly with the earliest review date set at least one year forward.What actually goes wrong, and whenFailureWhen it happensCostRedirects removed earlyRenewal, cleanup, handoverUndoes the migrationChained onto an earlier moveThe second or third migrationHits the 3-hop guidanceEverything sent to the homepageLaunch week, for speedDestroys the mappingOld URL inventory not exportedBefore launch, by omissionUnrecoverableNon-page assets forgottenAlwaysBreaks silentlyExport before you switchThe full URL list and the Search Console performance data. Afterwards, you cannot reconstruct them.One governance fix covers most of these: name an owner for the redirect map, review date a year out.
Most of them happen months after launch, which is why the redirect map needs a named owner and a review date. Source : Method, against the documented one-year retention (2026)

A sequence that works

Eight steps, ordered, with the documented figures attached where they exist.

Export everything first. The full URL list, the Search Console performance data by page and by query, and the current index coverage. Do this before touching anything, because it cannot be recovered later.

Build the mapping page by page. Old URL to its genuine equivalent. Where there is no equivalent, decide deliberately between a 410 and a redirect to the nearest real page, and record the decision.

Test the new site thoroughly before switching. The documentation puts this first for a reason: a migration onto a broken site compounds two problems.

Move all at once if you are small or medium. That is the documented recommendation, and it helps the index update faster.

Use 301 or 308, server-side. Not a JavaScript redirect, not a meta refresh.

Submit the change of address only if you are changing domain or subdomain. Not for protocol, not for www, not for paths.

Brief the business on the timeline before, not after. A few weeks or more for medium sites, longer for large ones, with the real comparison being the same period next quarter.

Then set the review date twelve months out. And write down who owns it.

An eight-step website migration sequence with documented figures attachedAn eight-step sequence for conducting a website migration, with the search platform’s documented figures attached at each point where one exists. The first step is to export everything before touching anything, comprising the complete address list, the search console performance data broken down by page and by query, and the current index coverage, on the basis that this material cannot be recovered once the old site is decommissioned. The second step is to build the address mapping page by page, from each old address to its genuine equivalent, and where no equivalent exists to decide deliberately between returning a four hundred and ten status code and redirecting to the nearest genuinely related page, recording the decision either way. The third step is to test the new site thoroughly before switching, which the documentation places first because migrating onto a defective site compounds two separate problems. The fourth step is to move all addresses simultaneously for small and medium sites, which is the documented recommendation and which helps the platform’s algorithms detect the move and update the index faster. The fifth step is to use server-side permanent redirects with the three hundred and one or three hundred and eight status codes, rather than a script-based redirect or a meta refresh. The sixth step is to submit the change of address notification only where the move is between domains or subdomains, and not for protocol changes, www normalisation or path restructuring within the same domain. The seventh step is to brief the business on the expected timeline before launch rather than afterwards, namely a few weeks or more for medium sites and longer for large ones, with the genuine comparison point being the equivalent period in the following quarter. The eighth step is to set the redirect map review date twelve months forward and to record who owns it.Eight steps, in order1. Export everything, before touching anythingURL list, Search Console by page and query, index coverage. Unrecoverable afterwards.2. Map page by pageNo equivalent? Choose 410 or nearest. Record it.3. Test the new site firstMigrating onto a broken site compounds two problems.4. Small or medium: all at onceThe documented recommendation.5. 301 or 308, server-sideNot JavaScript. Not a meta refresh.6. Change of address, only ifdomain or subdomain. Not protocol, www or paths.7. Brief the timeline before launchWeeks or more. Compare next quarter.8. Set the review date twelve months out, and write down who owns itBecause that is the documented period, and because nobody owns a redirect map by default.Step 1 is the only one you cannot do later. Everything else can be corrected; that one cannot.
Export first, because it cannot be recovered. Review date twelve months out, because that is the documented period. Source : Google Search Central, site move with URL changes (2026)

Where to go next

You want the technical requirements themselves. Technical SEO: three requirements.

You are changing hosting rather than URLs. B2B website hosting and performance.

You are working on speed. Website speed: what matters.

You are deciding whether to rebuild at all. Signs it is time for a B2B website redesign.

You are choosing a platform. The best website platform for B2B.

You want the measurement to survive the move. What GA4 does not show.

In short

  • Keep redirects at least one year. That is the documented period for signals to transfer, and considering indefinitely is the stated user-facing advice.
  • Use server-side permanent redirects, 301 or 308, and note that permanent redirects do not cause a loss in PageRank.
  • Chain limit: 10 supported, 3 or fewer advised, with fewer than 5 as the outer bound.
  • The change of address tool covers domain and subdomain moves. It is not needed for HTTP to HTTPS, www to non-www, or path changes.
  • Small and medium sites should move everything at once, which helps the index update faster.
  • Expect a few weeks or more of fluctuation on a medium site, longer on a large one. Brief that before launch.
  • A hosting change with no URL change is different guidance entirely, where the DNS change is the move and there are no redirects.
  • The commonest failure is removing redirects early, at a renewal or handover, inside the documented year.

Export the URL list today, before anything moves. Book a diagnostic, or see how we approach B2B websites.