Should you disavow a toxic backlink?

In most cases you should not. The disavow tool solves one narrow problem, and the anxiety-driven use of it, which consists of feeding it every domain a tool flagged red, is more likely to cost you links than to save you from anything. If you arrived here after seeing an audit full of red, the short answer fits in one sentence: do nothing for now.

What actually happens with spam links

Search engines have to handle the fact that anyone can link to anyone. If junk links dragged a site down, damaging a competitor would cost twenty euros and an afternoon. That is why the standard treatment is to ignore suspicious links rather than to punish the site receiving them.

So the pile of adult-site, casino and Cyrillic-directory links your audit flagged is, in all likelihood, already being discounted. It is neither counted in your favour nor held against you. Doing nothing is a defensible strategy and usually the correct one.

Disavow or do nothing: the table that settles it

The first three rows cover very nearly every real case. The last two are the only ones where the file has demonstrable value.

The two situations where it is justified

A manual action. Search Console reports one, naming unnatural inbound links. The message is explicit, dated, and it names the problem: there is no possible ambiguity, and nobody has ever had to guess whether they received one. Here the disavow file is part of the answer, alongside genuine removal requests and a reconsideration request that explains what changed.

Links you paid for. You ran a campaign, bought placements, and the sites will not take them down. Disavowing is how you formally step back from something you did, not from something that was done to you.

Outside those two, the honest answer is that a disavow file changes nothing you would ever measure. That is an uncomfortable thing to read when you have just paid for an audit, but the audit and the search engine are not looking at the same thing, and only one of them decides your rankings.

The file, line by line

The format is deliberately poor: a text file, one directive per line, comments prefixed with a hash.

# Links bought in 2024, removal refused by the publisher
http://example-a.com/partners.html
http://example-b.com/directory/companies/2024

# Whole network, no legitimate page
domain:link-network.example

Two rules govern everything else. A URL disavows one page, domain: disavows an entire site and all its subdomains. And the second form is irreversible in practice: it is the one that does the damage.

Why the file is dangerous

The syntax invites collateral damage. It is quicker to write domain:example.com than to list individual URLs, so people disavow entire sites, including the one good editorial link buried among the noise. A mediocre directory that also hosts a serious article disappears in a single line.

And the asymmetry is brutal. Removals take effect quickly, while getting a link counted again after you have disavowed it is slow and uncertain: you must delete the line, resubmit the file, then wait for the pages concerned to be recrawled, which can take months. So you are making a fast, irreversible-feeling decision on the strength of a third-party score that was never what Google uses.

If you genuinely have to do it

Five steps, in this order, and none of them is optional.

1. Confirm there is a manual action. Without a message in Search Console, stop here. The absence of a message is not a subtle signal, it is the answer.
2. List the links by hand. Not a tool's raw export: the list you could defend line by line in front of someone.
3. Ask for removals first. It is what counts most in a reconsideration request, because it demonstrates real effort.
4. Write the file using URLs, and reserve domain: for sites where no page is legitimate.
5. Keep a copy and the date. In two years nobody will remember what was disavowed or why, and that is exactly how a good link stays blocked while nobody understands why it does not count.

What a "toxicity" audit actually measures

No tool has access to what the search engine thinks of your links. A toxicity score is therefore an estimate built from signals visible from outside: the number of outbound links on the page, the topic, the age of the domain, the presence of spam-associated keywords, the share of marked links. It is honest work, but it is a deduction, not a measurement.

Two consequences worth keeping in mind. First, the threshold at which a link turns red is a commercial decision by the tool vendor: a tool that never finds anything does not sell. Second, a perfectly harmless link often ticks several of those signals. A personal blog abandoned three years ago, with little content and a domain about to expire, looks a lot like spam to an algorithm, and like nothing at all to a reader.

A link marked nofollow is not a toxic link either: it is a link that passes no authority, which is a different matter entirely. The difference is explained here, and confusing the two is the leading cause of pointless disavowing.

The negative SEO case

This is the fear behind most disavow files: a competitor buys a thousand links from dubious sites to drag you down. The mechanism exists, the offers exist, and yet documented cases of a serious site genuinely penalised this way remain rare.

The reason is structural: if it worked reliably, the whole web would be unmanageable, and treating suspicious links by ignoring them is precisely what prevents that. A sudden influx of dubious links to an otherwise normal site looks like what it is, and is handled as such.

What is worth doing, on the other hand, is knowing about it. An abnormal spike in your link profile is useful information, even if the recommended action remains, nine times out of ten, to do nothing and keep watching.

What to do instead

Spend the time on the links that matter to you. Know which ones exist, which are followed, what wording they use, and since when: a link that quietly turns into something else costs you more than a hundred spam domains that were already being ignored. That is a real change, measurable, and repairable by writing to someone.

If a pattern does worry you, look at your own anchor text distribution first. An over-optimised profile is a problem you created and can fix by dilution; someone else's spam usually is not.

And if you need to keep a record of what you decided about each link, including "seen, judged worthless, deliberately ignored", that is exactly the sort of thing a spreadsheet holds better than any tool. A judgement written down once saves you from making it again every year, and it is the only thing that stops the same audit producing the same panic every January. Write the reason next to the decision, not just the decision itself.