Off-Page SEO

Toxic backlinks: when to use disavow (and when to ignore)

Por Lucas ·

Objective criteria to decide on disavow without panicking. No guesswork, no fear-selling tools: just data and process.

Every week a client opens Ahrefs, sees a Domain Rating 2 site pointing at them with Russian anchor text, and messages in panic asking whether they need to disavow. Short answer in most cases: no. Google said publicly in 2024, through John Mueller, that the algorithm already ignores nearly everything third-party tools label as 'toxic'. But there is a narrow band of scenarios where disavow still saves the project, and that band is what this post targets. No pseudo-science 87/100 scores.

First, the vocabulary. 'Toxic score' does not exist in Google documentation: it is a proprietary metric from Semrush, Moz and Ahrefs, calculated with heuristics like suspicious TLD, no organic traffic on the source domain, repeated exact-match anchor. Useful as a first filter, dangerous as absolute truth. Before disavowing anything, you need an honest audit of your own profile; I recommend running the process described in How to audit a competitor's backlink profile adapted to your domain, then cross-referencing with Search Console: 7 underused reports and what to extract from them in the Links report to see what Google actually records.

When does disavow make sense? Three objective scenarios. One: you got a manual action in Search Console mentioning 'unnatural links to your site' - here disavow is not optional, it is mandatory for the reconsideration request. Two: you or a previous agency bought links on PBNs, ran guest post farms, or fired tier-2 spam at money pages - the history is documented and you want to clean up before the algorithm catches the pattern. Three: a real negative SEO attack, with thousands of links appearing in days, anchors carrying sensitive terms (casino, pharma, adult) that poison topical perception. Outside that, touching disavow.txt is almost always self-sabotage.

The process I use in practice at Lucas S.A. starts by exporting all referring domains from Search Console (not from Ahrefs, from GSC itself), filtering for domains with more than 5 links pointing in, and classifying in a spreadsheet: relevant, neutral, suspect, clearly spam. Suspect only becomes a disavow candidate when three signals converge: over-optimized anchor, source with no organic traffic in Semrush, and topical context fully off. Anchor alone is not enough - see Anchor text: natural distribution vs over-optimization for the tolerance band. And of course, before anything else, make sure your on-page is not bleeding authority through broken Canonical tags: common mistakes bleeding your organic traffic or sloppy 301 vs 302 Redirects: The Real Ranking Impact, because blaming external links when the problem is internal technical debt is the most common mistake I see.

Concrete numbers from a recent case: fashion e-commerce with 18,000 referring domains, 22% organic drop in three months. Client wanted to mass-disavow 4,000 domains flagged toxic by Semrush. We audited a sample of 200 manually: 87% were harmless scrapers, 9% old neutral directories, only 4% (roughly 160 projected domains in total) had a consistent PBN pattern. Disavowing those 160, plus fixing 47 broken canonicals and reviewing Smart interlinking: the internal authority map, recovered 19% of traffic in 90 days. The real win came from the technical work, not the disavow - but the disavow removed the ceiling.

Expensive mistakes I see repeatedly: domain-level disavow when the problem is a specific URL (use the domain: syntax only when the whole domain is bad), disavowing press links because they have 'click here' anchor (those are gold, see Digital PR for SEO: how to measure the real ROI of mentions), and re-uploading the file with each new audit without versioning - Google reads the last upload as absolute truth, so if you forgot a line, it disappears. Keep disavow.txt in git, with dated comments explaining why each domain entered. And never, ever, upload disavow generated automatically by a tool without human review of at least a 10% sample.

Practical takeaway: before thinking about disavow, spend two weeks solving what is under your control - technical, on-page, decaying content (see Content decay: spotting the posts quietly losing traffic). If after that you still have evidence of algorithmic penalty or documented spam pattern, then build a surgical disavow with at most a few hundred lines, versioned and justified. Disavow is not a broom, it is a scalpel. And for 80% of sites that never bought a link in their life, the best disavow.txt is the one that does not exist.

Nenhum comentário ainda

Seja o primeiro a comentar.

Deixe seu comentário

Entre com sua conta Canverly para comentar. Você pode usar a mesma conta em qualquer site da rede.

Entrar com Canverly