Skip to content
search-engine-optimizedby Netsolution

Migrating a site without losing its addresses

An address change comes down to four documented steps: redirect permanently on the server side, keep those redirects for a long time, rewrite internal links to the new addresses, and declare the new sitemap. Visibility then fluctuates, and that fluctuation is announced in advance.

What a migration actually moves

An address change does not move pages: it moves identifiers. Everything an engine has accumulated — what it crawled, what it indexed, the links pointing at you — is attached to addresses, and an address that changes becomes, to the engine, an unknown address.

Hence the only question that counts during a migration: can every old address say where it went? The rest — the new page template, the new platform, the new menu — belongs to the project, not to the migration.

The four documented steps

The engine vendor’s documentation lists few of them, and all are checkable.

The documentation recommends permanent server-side redirects, from the old addresses to the new ones, and cites the 301 and 308 codes.

officialSource : Google Search Central, “How to move a site”

Two words deserve picking out. “Server-side” first: a redirect written in JavaScript assumes the page is loaded and then run, which adds a step and an uncertainty where the HTTP protocol answers on its own. “Permanent” next, and this is the point handled in detail in the guide on canonicals: only the permanent form carries the signal that the target should become the address kept.

It asks that redirects be kept as long as possible, generally at least a year: that period is what allows all the signals to transfer to the new addresses.

officialSource : Google Search Central, “How to move a site”

The order of magnitude is interesting, but the justification is more so: the period is the one the transfer of signals takes. A redirect is therefore not a convenience for lost visitors, it is the channel through which the site’s past reaches its new addresses. Cut too early, it does not reconnect.

It also asks that the new site's internal links be changed, so that they point at the new addresses and not at the old ones.

officialSource : Google Search Central, “How to move a site”

This is the most often neglected step, because it is invisible: the site works perfectly with internal links that all go through a redirect. It simply tells a different story from the one the redirect asserts. A site whose internal links still name the old addresses is a site that hesitates.

The new sitemap is submitted in Search Console: this helps the engine discover the new addresses.

officialSource : Google Search Central, “How to move a site”

The verb used is “helps discover”, and it should be taken at its word: the sitemap speeds discovery up, it orders nothing. Declaring the new addresses is useful; believing it replaces the redirects is not.

Redirecting every old address to its final target, in a single hop

hypothesis

Effort
high
Time to effect
weeks
Risk
low
Durability
structural
A wall of numbered mail pigeonholes filled with letters and folders, in front of which an arrowed sign reads “Toques A B C D”.
Every addressee has a slot, and when the slot changes a sign says where it went. A migration does nothing else: the addresses move, and what decides the outcome is the sign you leave behind.Éric Messel — CC BY-SA 4.0 · Wikimedia Commons, relevé le

The inventory, done before and not after

A migration is prepared against a list of addresses, and that list is not improvised on switchover day. Three sources make it up, and they do not overlap: the addresses the site knows about itself, those the engine says it knows, and those third parties point at. The third is the easiest to forget and the most expensive to lose, since nobody but you will fix those links.

The expected result is a mapping from old address to new address, decided in advance, where every line points at an equivalent page and not at the home page. Redirecting en masse to the home page is a way of saying the inventory was never done.

The period afterwards, and the temptation to act

Content visibility may fluctuate temporarily during the move: the documentation calls this normal and states that positions settle over time.

officialSource : Google Search Central, “How to move a site”

That sentence is worth reading twice, because it cuts both ways. It protects against needless panic: it does not say every drop is temporary. What it licenses is waiting before concluding. What it does not license is explaining a lasting drop with “it is the migration” for months on end.

The practical distinction lies in what you watch. A fluctuation is read in positions and impressions. A migration error is read elsewhere: in the response codes of the old addresses, in multi-hop redirect chains, in newly unreachable pages. The first table calls for patience, the second calls for a fix — and it is consulted from day one.

What does not work

After a rebuild, redirect everything to the home page: Google will understand and redistribute the popularity of your old pages.

The documentation asks that old addresses be redirected to the new ones, which assumes a page-by-page mapping. A mass redirect to the home page establishes no equivalence of content, and nothing in the documentation promises the redistribution described here.

Two other reflexes are expensive. Changing addresses and content in the same move makes any diagnosis impossible: when two variables move together, no observation separates their effects, and that is exactly the situation a protocol exists to avoid. And dropping the old addresses from your inventory once the migration is “finished” means losing the one list that will let you check, six months later, that a forgotten redirect has not been dropped during a server update.

Sources

  1. How to move a site — Google Search Centralaccessed on
  2. Redirects and Google Search — Google Search Centralaccessed on
  3. Learn about sitemaps — Google Search Centralaccessed on

Block revalidated on

Frequently asked questions

How long should the old redirects be kept?
The documentation gives an explicit order of magnitude: as long as possible, and generally at least a year. The reason is written next to the period, it is the time the transfer of signals takes. Removing the redirects earlier interrupts that transfer.
Do you really have to rewrite internal links?
Yes, and it is an explicit request in the documentation. An internal link that goes through a redirect works for the visitor, but it makes the site say the old address is still the right one. The site then contradicts itself, which weakens the signal.
Is the drop observed after a migration abnormal?
A temporary fluctuation in visibility is announced by the documentation itself, which calls it normal. That does not mean every drop is normal: it means a drop over a few days proves nothing yet, and acting immediately often backfires.
Can you migrate in stages rather than all at once?
The documentation forbids nothing of the sort and splitting reduces the human risk. It does lengthen the period where two sets of addresses coexist, and therefore the period where signals are split. It is a trade between execution safety and how long the uncertainty lasts.