Disavow file 2026: when does it make sense, and how does it work?

What a disavow file is, and what it actually does

A disavow file is a plain text file that website owners can upload in Google Search Console. It lists URLs or entire domains whose incoming links Google should disregard when assessing that website.

What the file does not do matters just as much. It removes no link from the linking website. The link stays where it is, stays clickable and keeps bringing visitors. The file merely tells Google that certain links should not be taken into account when assessing your own site.

The verb “to disavow” means refusing to acknowledge something as your own. The word is easily misread as meaning that Google removes a negative signal. What actually happens is different: the link is passed over during assessment. We therefore mostly write “declare invalid” or “have disregarded” in what follows.

The tool was created in 2012, at a time when Google acted far more forcefully against unnatural and manipulative links, both algorithmically and through manual actions. It gave webmasters their first way of telling Google explicitly which links should not be taken into account.

With Penguin 4.0 in September 2016, the assessment of manipulative links became part of the core algorithm. Google stated that many spam links can now be recognised and neutralised automatically. That does not mean the disavow tool has become obsolete: Google still recommends it for substantial quantities of problematic links and for existing or likely manual actions.

When Google recommends a disavow file

The help page is unusually unambiguous. Google names two conditions that must both apply at the same time. First, a considerable number of spammy, artificial or low-quality links point at the site. Second, those links have caused a manual action or are likely to cause one.

By manual action Google means an intervention by a human being: a member of the spam team has reviewed the site and imposed a sanction. Affected owners find the notice in Search Console under “Security issues and manual actions”. A missing entry argues against a manual action already having been imposed. Google does, however, explicitly name cases in which problematic links are likely to lead to a manual action.

In most cases Google can assess which links to trust without additional guidance. Most websites will therefore not need this tool.

In the same help document Google calls the disavow tool an advanced feature and warns that using it incorrectly can harm a site’s performance in Google Search. In addition, John Mueller, Search Advocate at Google and responsible for communication with site owners, has repeatedly said that nothing needs to be done about spammy backlinks because Google ignores them anyway; on comment spam he stated explicitly that it affects rankings neither positively nor negatively.

What the spam policies say — and what they do not

Link spam is described there at length: bought and sold links, excessive link exchanges, automatically generated links, paid guest posts with optimised anchor text, and widely distributed links in footers and page templates. Every one of these descriptions is written from the perspective of whoever places such links. Google states no general rule under which arbitrary low-quality links from third parties automatically lead to a penalty for the target website. Under certain circumstances, incoming links can nevertheless influence how a website is assessed, particularly where they form part of artificial or manipulative link patterns.

As examples of such patterns Google names paid links, link exchange schemes and artificially built links; the help page for the disavow tool presupposes exactly that case. The difference therefore lies in the order of magnitude: what is described are interventions against substantial, recognisably organised link patterns, not against individual low-quality references.

What the spam score of backlink tools really measures

Almost every backlink tool displays a number between 0 and 100 for each link and calls it a spam score or toxicity score. The presentation suggests it is an assessment of harmfulness. In reality it is the result of a checklist, and with several providers that checklist is publicly documented.

For DataForSEO, the data supplier behind numerous tools and behind Rankmio, the calculation is fully documented. Eighteen signals are examined on the linking page itself; each adds or subtracts a fixed percentage, and the sum is the spam score displayed. For an entire domain, the average across all examined pages is used.

Every one of these characteristics describes the craftsmanship of a web page: its encryption, its markup, the length of its name, the number of its references. About the intent behind a link, and about Google’s assessment of that link, the list contains no information.

The score does not say whether Google devalues the link. It says whether the page was built tidily.

Two kinds of error follow from this. Dated but entirely harmless pages receive high scores: a club website maintained for years without HTTPS, without social media tags and without a contact page reaches seventy points, even though the link from it was placed editorially. Conversely, professionally built pages from link-selling networks receive low scores, provided their operators know the basics of the craft.

Other providers calculate differently but follow the same approach in principle. Moz states it evaluates twenty-seven signals, Semrush more than fifty; both are disclosed only in summary form, whereas DataForSEO names all eighteen signals together with their point values.

Two providers take a methodologically different route, and both strike us as sounder. At Majestic, Trust Flow measures how close a domain sits, within the network of links, to a set of trusted seed pages, while Citation Flow merely reflects the quantity of incoming links; the ratio between the two is considered a usable indicator of link networks, because those generate a great deal of quantity at little trust. Sistrix bases its judgement of a linking domain partly on whether it is affected by an algorithmic sanction and how well it is itself linked, measured against the visibility index. Data about a domain’s actual performance therefore enters the calculation, not just characteristics of its source code. There is, however, no published formula, so the quality of the method cannot be verified from outside.

Pattern detection instead of a single score

If a single metric cannot carry the decision, the alternative lies in observing characteristics that coincide. One conspicuous trait can be put down to chance; when several conspicuous traits occur at the same time, the likelihood of a common cause increases.

The example is a real link profile we analysed with the owner’s consent. It belongs to a small blog about angling. Nine different domains link to this website. All nine use the same anchor text — the clickable text carrying the link — running to ninety-two characters, and that text openly advertises the sale of backlinks. All nine links sit on pages with the same URL pattern /all/3320/32.html. All nine were first seen between 11 August and 1 September. And all nine are dofollow links, passing on ranking weight under the default behaviour.

The simultaneous occurrence of several conspicuous traits argues clearly against a chance origin. The domain ratings of these nine pages, by contrast, range from 1 to 216: anyone filtering by that widely used metric would have missed roughly half the network.

Reliable individual signals

The follow status. A link carrying the nofollow attribute passes on no ranking weight under the usual treatment, so declaring it invalid has as a rule no effect. Since March 2020, however, Google treats nofollow, ugc and sponsored as hints rather than binding directives, so the status offers no certainty. For our own pre-selection within a link profile the follow status can therefore be a sensible exclusion criterion — that is Rankmio methodology, not a Google rule.

The anchor text is the most telling individual signal, though not in the form in which it is usually examined. What matters is not how often a commercially optimised anchor text occurs, but its identical repetition across several unrelated domains. A single optimised anchor text appears in every organically grown profile. The simultaneous occurrence of identical text blocks on several unrelated domains is a strong indication that the links were placed in a coordinated way — an indication, not proof, since franchise systems, partner programmes and central copy templates produce such repetitions as well.

The URL pattern of the linking pages gives networks away, because they are generated from templates. Where the same path pattern repeats across several domains, that argues for the pages belonging to one group.

The first-seen date serves to group links, not to judge them. A cluster of new links within a few days points to a campaign. A link’s age is not a seal of quality, and youth is not a defect.

Conspicuous but harmless cases

The same example profile contained thirteen links with a spam score of fifty. They came from URL shorteners and analysis tools that leave their result pages publicly accessible after someone entered the domain into a form once. All thirteen carried the same job number in the URL path and were marked nofollow. Under the usual treatment they pass on no weight, and no harm can be inferred from them.

Comparable caution applies to business directories with a dated appearance and genuine visitor numbers, to pages that copied content and carried the embedded link along with it, and to compromised websites that serve spam temporarily and then recover.

One further case is rarely considered: links from domains owned by the same operator. In the example profile, twelve of thirty-six links came from two further websites belonging to the same person. Such links change the picture of a profile considerably. Shared ownership of a domain is not on its own sufficient grounds for exclusion; links should not be declared invalid merely because of shared operator structures.

Creating a disavow file: format, example and upload

Should a file genuinely be required, the formal requirements are tight. Google rejects a malformed file immediately and names the errors it found.

• It must be a text file with the extension .txt, encoded in UTF-8 or 7-bit ASCII.
• Each line holds either a single URL or an entire domain. Two entries on one line are not permitted.
• An entire domain is marked with the prefix domain:, written as domain:example.com — without protocol and without www. domain:example.com covers the whole domain including its subdomains; domain:blog.example.com covers only that one subdomain.
• Subdirectories such as example.com/blog/ cannot be specified; Google’s help states this explicitly.
• Lines beginning with a hash # are ignored by Google and are suitable for dates and reasons.
• The file may contain at most 100,000 lines including blank and comment lines, and must not exceed 2 MB; individual URLs are limited to 2,048 characters.

A properly documented file can therefore look like this:

# Link-selling network, identical anchor text, identified on 22/09/2026
domain:fashionclothingnews.com
domain:uncledspizza.com
domain:plurio.shop

# Single page instead of the whole domain,
# because the rest of the site is a genuine business
https://example.com/compromised-page

Three properties of the procedure cause recurring misunderstandings. First, every upload replaces the existing list for that property completely; it is not an addition. Second, the stored list cannot be downloaded — the interface only shows how many pages and domains are covered. Third, the effect sets in only once Google recrawls the pages concerned, which according to Google takes several weeks; reversing it takes the same amount of time.

The type of property matters as well. The disavow tool does not support domain properties. The file has to be uploaded for a URL-prefix property, for example https://www.example.com. Anyone whose site exists in Search Console only as a domain property needs to add a URL-prefix property for the upload. The list then applies to that property and the addresses beneath it.

Common mistakes in practice

Declaring an entire domain invalid when it runs a genuine business of its own. A line with the domain: prefix covers every link from that domain, editorial ones included. For press distributors, business directories and news portals, listing the individual URL is preferable.

Uploading an incomplete list. Because the upload replaces the previous file entirely, every domain missing from the new file counts as acknowledged again. Anyone wanting to add entries must upload the old ones together with the new.

Selecting wholesale by a single metric. Anyone excluding every link below a certain domain rating regularly catches editorial mentions and smaller specialist blogs. The mistake is typically noticed weeks later, through changed rankings.

Formatting errors in the domain: line. The spelling domain:https://www.example.com/ is invalid; the correct form is domain:example.com.

Excluding platforms with many independent users. A line such as domain:medium.com covers every link from that platform, including an editorial piece about your own company.

The situation at Bing

Microsoft removed the disavow feature from Bing Webmaster Tools in October 2023, together with its API. The company argued that its own systems can now distinguish natural from unnatural links by themselves. A disavow file created for Google has no effect on Bing in any case.

For visibility in AI-driven systems this matters, because both ChatGPT and Copilot draw on the Bing index. Anyone examining their link profile with generative search systems in mind has no corresponding lever there — and, according to Microsoft, does not need one.

A methodical approach

The first step is checking Search Console for a manual action over unnatural links. If there is none, then according to Google’s documented position the probably correct step is to do nothing.

Next, establish whether a disavow file already exists for the property. On inherited projects this is the most common blind spot, because a previously commissioned agency may have entered domains the new caretaker would rather keep.

The assessment itself should address groups rather than individual lines. A network of nine domains represents one decision, not nine. Each group should be documented with a comment line so that its reason remains traceable much later.

Before settling on a final selection, compare the candidates against your own traffic data. Demonstrable referral traffic argues against a hasty exclusion and should be taken into account. On its own, however, it does not prove that a link arose naturally — paid advertorials, affiliate pages and partner networks bring visitors too.

Finally, it is worth estimating the expected effect. In the example profile, even a fully effective exclusion would have changed only a handful of clicks per month, because the site recorded eleven clicks in four weeks in total. An order of magnitude like that belongs in any sound piece of advice.

What this means for Rankmio

Rankmio continues to use the spam score as one signal, not as an automatic verdict. Alongside it we will state that the score describes the technical condition of the linking page and is not an assessment by Google.

In addition, conspicuous patterns are made visible: identical anchor texts, recurring URL patterns and clusters in time. Rankmio does not tell you that a link is toxic; Rankmio shows you why a link stands out and what pattern lies behind it. The aim is not an automatic list of toxic links, but a traceable basis for the owner’s decision.

The planned tool creates the file on request and points out the possible consequences beforehand. Its use follows no rules published by Google and happens at your own responsibility.

Frequently asked questions about the disavow file

1. What is a disavow file and what does it do?

   A disavow file is a text file you upload in Google Search Console to tell Google that certain incoming links should not be taken into account when assessing your website. It lists one URL or one domain per line, the latter with the prefix “domain:”. It removes no link from the linking website; the link remains in place and keeps bringing visitors.

2. When does Google recommend the disavow tool?

   Google names two conditions that must both apply: you have a considerable number of spammy, artificial or low-quality links pointing at your site, and those links have caused a manual action or are likely to cause one. Google states plainly that in most cases it can assess which links to trust without additional guidance, so most sites will never need the tool.

3. Should I disavow every toxic backlink?

   No. A high spam or toxicity score from an SEO tool is not on its own sufficient grounds for a disavow file. Google recommends the tool only under considerably narrower conditions, and the score describes the technical condition of the linking page, not Google’s assessment of the link.

4. Can a disavow file harm my rankings?

   Yes, an incorrectly assembled file can do harm. Google calls the tool an advanced feature and warns explicitly that using it incorrectly can harm a site’s performance in Google Search. Anyone who excludes editorial links loses their effect, and reversing the decision again takes several weeks.

5. Do I have to get bad backlinks removed?

   Not automatically. Google recommends having problematic links removed from the linking websites where possible. The disavow feature is intended for cases where that is not possible and the stated conditions are met.

6. How do I create a disavow file?

   Create a text file with the extension .txt in UTF-8 or 7-bit ASCII and write one URL or one domain per line. An entire domain is written as domain:example.com, without protocol and without www; that covers all its subdomains as well. Subdirectories cannot be specified. Lines beginning with a hash are comments and are suitable for dates and reasons. The file is then uploaded in the disavow tool of Search Console. Important: the tool does not support domain properties; the file has to be uploaded for a URL-prefix property such as https://www.example.com.

7. How do I update an existing disavow file, and when does it take effect?

   Every upload replaces the previous file completely. Anyone wanting to add entries must upload the old ones together with the new, otherwise the missing domains are silently acknowledged again. The list applies to the property it was uploaded for and to the addresses beneath it. The effect sets in only once Google recrawls the pages concerned; according to Google that takes several weeks, and reversing it takes the same amount of time.

8. What does the spam score of a backlink tool measure?

   In the case of DataForSEO it measures eighteen technical signals on the linking page itself. Running on HTTP instead of HTTPS adds thirty points, missing verification tags add twenty, a suspicious top-level domain adds twenty, and a domain name shorter than three or longer than twenty characters adds ten. Contact details and social media meta tags subtract points. The score describes the craftsmanship of a page, not Google’s judgement of the link.

9. How do you recognise a link network?

   Not by a single number but by coinciding patterns. The reliable signals are an identical anchor text across several unrelated domains, an identical URL pattern on the linking pages, a first-seen date clustered within a few days, and the dofollow attribute on all of them. The simultaneous occurrence of several such traits is a strong indication of coordinated placement, but not proof.

10. Can incoming links from third parties harm my site?

    Google states no general rule under which arbitrary low-quality links from third parties automatically lead to a penalty for the target website. Under certain circumstances, incoming links can nevertheless influence how a website is assessed, particularly where they form part of artificial or manipulative link patterns; as examples Google names paid links, link exchange schemes and artificially built links. John Mueller has said repeatedly that Google ignores spammy backlinks in many cases and that comment spam affects rankings neither positively nor negatively.

11. Does Bing have a disavow tool?

    Not any more. Microsoft removed the disavow feature from Bing Webmaster Tools in October 2023, arguing that its systems can now distinguish natural from unnatural links on their own. A Google disavow file has no effect on Bing in any case.

Sources

• Google Search Console Help: Disavow links to your site
• Google: Spam policies for Google web search
• Google Search Central: Evolving “nofollow” — new ways to identify the nature of links
• DataForSEO: What is Spam Score and how is it calculated
• Bing Webmaster Blog: Removal of the disavow feature in October 2023
• Sistrix: How can I determine the quality of a potential backlink page?