What happens to your links when you change domain

There is a domain sitting in my registrar account with nothing on it. No site, no pages, no mailbox. It costs about 14 dollars a year and the only thing it does is forward requests to an address somewhere else.

That renewal is the cheapest line on my infrastructure bill and the one with the most hanging off it. If it ever fails, every link anyone ever pointed at that domain stops working in the same hour.

I buy and sell placements and I keep a ledger of both sides, with the destination URL in a column. So when a site of mine changes address, I can put a number on the exposure instead of a mood. I open the file and count rows.

A link is a string on somebody else’s server

Worth stating plainly, because the word “link” makes it sound like a connection between two live things.

It is a piece of text in a database you have no access to, written by somebody who has probably forgotten writing it, and that text is your old address. Nothing follows you when you move. No registry, no forwarding table, no notification anybody receives.

Change domain and you have not touched a single one of those strings. You have changed what they resolve to, everywhere at once, silently.

The only thing standing between that and a total write off is a rule running on a server you rent.

What survives if you do the work

Most of it.

Map each old address to the page on the new domain that genuinely replaces it, serve a permanent redirect for every one, and the bulk of what those links were worth lands where you want it. Search engines have been handling this for twenty years and they handle it well.

I am deliberately staying out of redirect mechanics here. Chain length, permanent against temporary, what each extra hop costs you. That has its own piece, and none of it is where migrations actually fail.

Migrations fail on the map.

Everything to the home page

The fastest possible migration is one rule sending the entire old domain to the new front page.

It passes every check. Nothing 404s. Whoever ran it can report a clean move.

What it destroys is specificity, and specificity was the whole value. Somebody wrote about one narrow subject, linked to your page on that subject with an anchor describing it, and the result is a recommendation about a thing. Point it at your front page and you have converted that into a recommendation about a company.

Search engines have treated wholesale home page redirects as a soft failure for years, so the consolidation people imagine they are buying never shows up either. What shows up is a dead end returning a healthy status code, which nothing in your monitoring will flag.

An incomplete map

Quieter, and I run into it more often.

Most maps get built off the sitemap or a CMS export. Both describe the pages you currently have. Neither describes the pages other people linked to, and those two sets stopped matching years ago.

The gaps are always the same kinds of thing:

• addresses from a URL structure you replaced in 2021
• pages you deleted, which somebody had already cited
• category and tag paths nobody thinks of as pages at all
• an address with a tracking parameter still stuck to it, copied out of a browser bar and pasted into an article

Every one you miss is a link that survived the move and then landed on an error.

Build the map from the links

So I build it in the other direction.

Pull every address on the old domain that anything external points at. My own placed and received ledgers come first, because those are the entries with invoices behind them. Then a backlink tool for everything I did not arrange myself. Then the access logs from the old domain, which name the addresses that real crawlers and real people were requesting.

Deduplicate into one column and that becomes the mandatory list. The sitemap goes on afterwards as the nice to have.

The ordering matters because it puts the effort where the money is. An unlinked page missing from the map costs you a trickle of traffic. A linked page missing from the map costs you an asset somebody paid for.

The pages with no equivalent

Every move has a set of these and they get handled last, badly.

A page with links pointing at it and nothing on the new domain doing the same job. It was merged, or the section got dropped, or the product it described is discontinued.

It still needs a decision. Send it to the nearest page that answers the same question, which is usually a parent category and almost never the home page. Or rebuild it on the new domain, which is worth the writing time when the link pointing at it is good enough. Or let it return an error and take the loss honestly, because a redirect you cannot defend out loud is worse than an error page telling the truth.

Leaving it off the map is the only option here that is not a decision. An unmapped address with links on it behaves exactly like a deleted page with links on it, and losing published pages has its own piece.

The renewal is a business dependency now

Here is the model shift I would push hardest.

A migration feels finite. Plan it, run it, watch the numbers for a month, close the ticket.

The redirect never closes. It is permanent infrastructure with an annual invoice, sitting next to your DNS and your hosting, and it stays that way for as long as any of those links matter to you. For a good link, that is longer than most companies stay in business.

So the old domain belongs on the list of things that cannot be allowed to fail. Auto renew enabled, a card that is not about to expire, an account more than one person can get into. I keep that renewal date beside my server payments because the failure mode is identical: a card declines, a warning lands in a mailbox nobody reads, something goes dark.

I came close to exactly that. A card I had replaced months earlier was still on file at one registrar, and I found out from an expiry notice I happened to read on a weekend. There was no website on that domain, so it had never once registered in my head as critical.

If you bought the links, price this properly

Paid placements make the arithmetic uncomfortable.

The money is spent the day the article publishes. What you hold afterwards is a string on a page you do not control. You cannot edit it, and asking the seller to change it means asking a favour on something already paid for, which is a bad way to spend a working relationship.

After a migration, all of that inventory is only live because a rule you maintain says so. The links did not move. A machine you rent repoints them on every request, indefinitely.

Put numbers against it. A few thousand across dozens of placements over two years, and the entire ledger now depends on a domain renewal costing less than lunch. That ratio is why I check the placed links file by hand after any move instead of trusting a crawl. A crawler will confirm that every address resolves while half your paid anchors quietly arrive at a page nobody bought.

I have been on the other end of it as well. A site carrying one of my placements changed address, I found out from my own monthly crawl rather than from a person, and the link somebody had paid real money for was landing on a front page.

When I think a move is worth it

The unpopular half.

Most domain changes I have watched were optional. A rebrand somebody wanted. A shorter name. The matching dot com finally freeing up after years of waiting. Real motivations, and every one of them discretionary.

The ones that earn the risk are the ones where staying put is the expensive choice. A trademark problem. A history attached to the domain that you inherited and cannot clean up. A name that costs you a deal every week because nobody can spell it over the phone.

Moving because the new name reads better on a pitch deck means betting an asset you spent years and real money assembling against a preference. On a site with a serious link profile I would not take that. On sites with no profile to lose I have taken it more than once, and that is a different bet wearing the same clothes.

What I have not measured

I cannot show you a clean before and after from a migration of my own. Reporting across my portfolio is in a state I do not trust, so any position or session figure I quoted would be decoration on a claim I had never verified.

What I can show is what the ledger held going in and what every destination returned coming out. That is narrower than what you probably wanted, and it is the part I will defend.

The migration map template I work from, and the checks I run against it before switching anything over, are here.