Disavow files and when they are a mistake

Nothing useful happens on screen when you upload a disavow file. You are told the file was accepted. No list of affected links, no date it takes effect, no readout afterwards saying what it did.

I have never uploaded one. That is not a principle I arrived at cleanly. I got about an hour into building a file on a domain a backlinks vendor had scored at 48% spam, then opened fifty of the flagged domains in a browser and found scraper mirrors, directory clones and sites republishing an RSS feed with the links left intact. The file went in the bin.

I buy placements and I sell them, and I pay roughly two cents a lookup to vet domains before money moves. How that spam percentage is assembled, and why my own site scored worse than every seller I was checking at the time, is a separate piece. This one is about the decision you make afterwards, which is where the damage lives.

Almost everything else you do is a request

You request indexing. You request reconsideration. You ask a seller for a placement and they decide.

A disavow file works differently. It is a text file, one domain or one URL per line, uploaded against a specific property. It is an instruction to discount those links and it is taken at face value. Nobody reads your reasoning. No human is involved on the other side at all unless you already have a manual action open.

The only judgement applied anywhere in that chain is yours. Made from a dashboard, usually in one sitting, usually while mildly worried about something you cannot name.

Reversible on paper

Delete a line, upload the new version, and people will tell you the whole thing was reversible.

Taking a domain out of the file makes the link eligible again. It does not put back what that link was doing. The page has to be recrawled and reassessed, whatever it carried has to work its way through again, and there is no published timeline for any part of that. Weeks is the guess everybody repeats. It is a guess.

So what you have is a decision you can undo in the interface and cannot undo in practice inside any window you would plan around.

The number you disavowed over will not move

This is the part that surprises people the first time they do it.

A disavow changes nothing about the link. The page is still up, the anchor is still sitting in the paragraph, and the link is still on the open web where every crawler can see it.

The vendor carries on counting it. Pull the same report next month and the percentage is identical, or higher, because two more scrapers found you in the meantime.

You uploaded a file to make a number less frightening and the number is untouched. The common response at that point is a second, larger file.

Two different questions

A spam score answers one question: does this domain look like the sort of domain that tends to be junk. Thin pages wrapped around link blocks, enormous outbound counts, exact match anchors repeating across a network. That is a real signal and I pay for it every month.

The question you need answered before you disavow is a different one. Is this specific link, on this specific page, currently carrying anything toward my site.

No vendor answers the second question. None of them claim to. The two get collapsed because they arrive in the same table, three pixels apart, and one of them has a percent sign on it.

A domain can fail every outbound test in that model while still hosting the one link an editor placed on purpose because your page was worth citing. That link sits in the same red row as a mirror of your own article.

Ninety seconds and a browser tab separates those two cases. A dashboard never will.

The failure is silent, then it is smeared

Say you disavow something that was quietly helping. Here is what that looks like from where you sit.

No notification arrives. The effect does not land as a step either, because it applies as pages get recrawled, so it spreads itself across weeks.

A step you can attribute. Something falls on Tuesday, you look at what you shipped on Monday, you have a suspect. A smear hands you nothing. By the time the shape of it is visible you have published four articles, an update has rolled out, and a competitor has done something you never noticed.

So the decline gets pinned on one of those. The file never comes up, because it left no trace and you had already filed it under housekeeping.

Price the bet

If you are right and those links were junk, Google’s stated position for years has been that most of that gets discounted already, without being asked. Being right buys you very little.

If you are wrong, you have removed links you paid for or earned, and you get them back on a schedule nobody will put in writing.

Small upside, real downside, and no instrument that measures either one. There is also the afternoon it costs. Building the file properly means opening every domain that goes into it, and people who actually do that tend to stop wanting to upload it.

The two cases that earn it

A manual action you have received. Not one you fear. A message in Search Console naming unnatural links to your site. Now there is a defined scope, a specific claim from the search engine, and a reviewer on the other end who will read what you did.

A negative campaign you can document. Dated evidence, thousands of referring domains appearing inside a window you can point at, a ramp you can put on a chart, and an honest account of why none of them are yours. Even then, removal requests go first. The file is what you are left holding after those fail.

I will grant a third that is closer to bookkeeping. You bought links at scale, you still have the list, and you are cleaning up after yourself. You know what belongs in the file because you have the invoices.

The three that do not

A spam score on a dashboard. That is a screen. Screens throw false positives by design, because the alternative is missing the real ones.

A domain you do not recognise. Not recognising a domain is the normal condition of owning a website. You did not ask for most of what points at you, and a scraper finding your post is not an event.

Unease about a profile you inherited. Buying a site and immediately disavowing what came with it is the most expensive version of this mistake, because that profile is a large part of what you just paid for.

If you build one anyway

Two rules, and I would hold myself to both.

Never use the domain: prefix on anything you have not reviewed by hand. That line removes every link from the host, including any link it might send you next year. A URL line removes one. The tool nudges you toward the domain form because it looks tidier on the page, and tidier is the wrong instinct here.

Write the reason beside every line in your own copy, with the date. If you cannot put down one sentence explaining why that particular link is hurting you, the line comes out.

That second rule on its own would empty most of the files I have seen.

Deleting assets you never valued

Here is the position I will argue about with anyone.

Uploading a disavow file because a tool coloured some rows red is a decision to delete assets you have never once valued.

You would not delete two hundred pages off your own site because a plugin flagged them, without opening a single one to see what was on it. This is the same act. The pages happen to sit on somebody else’s server, so you never watch the deletion happen and nothing appears in your own logs.

I cannot prove the file I nearly uploaded would have hurt me. I did not upload it, so there is no counterfactual on my desk and there never will be. Nobody who did upload one has a counterfactual either, which is the real reason this practice never gets tested by anybody.

What I do instead is open the domains, look at my own outbound surface, and write the number in a row with a date on it. The tool has been available since 2012. It will still be there in a month, and the file you do not upload today costs you nothing. The rest of what I have worked out from both sides of these deals is here.