What a redirect actually does to a link
There is a 302 on one of my own sites that I cannot remember creating.
Temporary redirect. Meaning: the real page is still at the old address, we are only diverting you for now, do not update anything. It has been saying that for long enough that I have no memory of the migration it was put there to cover.
That is the whole problem in one artefact. Nothing about it is broken. It resolves, and the chain ends in a 200. No error state, no alert, no line in any report. It is quietly telling everything on the other side something that stopped being true years ago.
I buy link placements and I sell them, and I keep written ledgers of both directions. Redirects are the failure mode in those ledgers I trust least, because it is the only one that looks like success.
The two things a redirect can say
A 301 says: this address is now that address, permanently, update your records. Links pointing at the old url get read as pointing at the new one. Most of what the old page had comes across with it.
Most. I will come back to that word.
A 302 says: the real page is still here, we are diverting you for the moment, keep the old address on file. Every link aimed at the old url stays credited to a url that no longer serves anything.
From a browser the two are identical. You type the old address, the new page appears. The difference lives in a single header nobody opens. If you want to see it without a terminal, load the old url with the network tab open and read the first row rather than the last one.
The temporary one that outlives everybody
Somebody picks a 302 because it is the cautious option mid migration. Two weeks, then we make it permanent. The launch goes fine and nobody thinks about it again.
Three years on, the redirect is still temporary. Officially the page has not moved. Everything upstream has been told, repeatedly, not to commit to the new address.
I do not think this happens because people are careless. It happens because a wrong 302 and a correct 301 produce exactly the same experience for every human in the loop. Nothing in the feedback ever points at it.
If the move is permanent, say permanent. A 302 left up for two years is a false statement made to a machine with no way of checking.
Chains, and the part I cannot measure
A chain builds by accumulation rather than by decision. Page moves in 2022, folder gets restructured in 2024, domain changes after that, and the original link now runs three rules deep.
Everyone says value leaks at every hop. I have repeated that line myself. I have never measured it on anything I own, and I am not going to publish a number I invented.
Two things I can stand behind. Each hop is a full round trip before any byte of the real page arrives, and I test pages over live carrier lines because I run mobile proxy infrastructure, so I have watched three hops become a visible delay on weak signal.
The other is that chains break from the middle. Somebody tidies a redirect table two years later, deletes the rule that looked orphaned, and everything above it now terminates in a 404. A link that survived two migrations dies in a cleanup.
So flatten them. Point the original address straight at the current page. Five minutes of work, and it removes something a well intentioned person will otherwise delete.
Where it lands is the whole thing
Status code and hop count are what people check. Neither is the one that costs money.
A redirect can be permanent, single hop, 200 at the end, clean in every tool you own, and land on a page with nothing to do with what the link was about.
The everyday version is consolidation. A product page gets folded into its category. The item page is gone and everything pointing at it arrives at a list of forty other items. Somebody clicked for one thing and got a shelf.
The blunt version is a single rule sending every dead url to the home page. I have argued elsewhere that this is dishonest to the reader who clicked. The mechanical half is that a redirect to a page which does not answer the original question gets treated as a deletion anyway. You did not rescue the page. You hid the deletion from your own 404 count.
A redirect landing somewhere unrelated is a delete with better manners.
What happens to a placement you paid for
I sell dofollow placements on my own sites for $250. That is roughly what is at stake per link, and it is why this section exists at all.
You buy a link inside an article. It goes live, you log it. Eight months later the seller merges two thin guides covering the same subject into one better guide, and the weaker url gets a permanent redirect to the survivor.
Good housekeeping. I would do the same. But your paragraph sat in one of those two articles, and if it was the one that got absorbed, it probably did not survive the edit. What you hold now is a redirect to a page you were never on.
Nothing announces this. The seller is usually not being dishonest either. They were improving their site and they had no list of who had links inside what.
I know they had no list, because mine only started in July of this year. Before that I merged and pruned pages on my own properties without checking whether anybody had paid to appear in them. I cannot tell you how many placements I broke doing it. There was no record to break them against.
The four checks
Request the url without following anything automatically, and print every hop. curl hands you each status and each location header in order.
Read the status of every hop as you go. A 302 that has been sitting in a chain for years is a finding by itself.
Open the final url and read it like a person. Is this page about the thing your link was about. No tool answers that question. Fifteen seconds, and it catches the failure everything else waves through.
Search the destination page for your own url. The redirect told you the page moved. It never told you your link came with it.
That fourth one matters most and gets skipped most. The full post purchase check, covering anchor drift and the rel attribute, is a separate job. This is the redirect shaped slice of it.
Two lines in an email
Before money moves, one sentence they have to answer:
If this article gets merged or moved, does my link move with it, and will you tell me.
Then, if you can get it: no blanket redirect of my placement into a hub or a category page.
Most sellers will agree to both and most will forget. Enforcement was never the point. You get something written down to reference, and you learn something from how they answered. Anyone who tells you it never happens here has not run a site through a restructure.
A redirected placement is dead until I have read the destination
Not flagged. Not pending review. Out of the count entirely, filed next to the 404s.
People find that harsh because the link technically resolves. Resolving is precisely why nobody looked at it. A dead bought link gets chased inside a week because it is loud. A redirect into a category page sits there for a year with everyone satisfied.
In practice it is one extra column in the ledger. Last destination read, with a date beside it. Blank cell, not counted, whatever the crawler says.
What I cannot tell you
How much a 301 actually passes. Every figure in circulation traces back to something someone said in some year, and they do not agree with each other. I have not run an experiment on my own properties clean enough to add to that argument.
What I can tell you is the gap between a link arriving somewhere relevant and a link arriving somewhere unrelated. I have watched that on real pages and it is not a rounding error.
A move, not a fix
A redirect tells everything on the other side where the page went. That is the entire function. It does not undelete anything and it cannot manufacture relevance.
So if you cannot name the destination and say out loud what it covers, you have not moved that page. You deleted it and left a forwarding address to an empty room.
On my own sites the routine is dull now. Before anything gets merged or removed I check the ledger for who has a link inside it, pick a destination that answers the same question, and where there is no such destination I leave the 404 standing and take the hit. A loud problem gets fixed. A quiet one just gets older.
The commands I use to walk a chain, and the ledger columns sitting behind all of this, are here.