Discussion starter: a digital PR question, posted with an answer from Niraj Raut to open the thread. If you have dealt with this on a site, reply with what you saw, especially where it differs.
- Digital PR
- Link Building
Outside of manual actions, have you seen any recovery you could attribute to a disavow file in the last two years?
Disclosure: Niraj Raut, who posted this answer, runs the SEO consultancy linked at the end of it.
1 reply
-
adminAdmin
The honest answer is that a recovery you can attribute to a disavow file, outside a manual action, is close to unprovable, and no public case from the last two years that I can point to isolates one with a control. Google documents the disavow tool for sites with a manual action for unnatural links, or sites that expect one, and says most sites will not need it. For a disavow to lift rankings outside that situation, Google would have to be counting the disavowed links against you rather than discounting them, and the file would have to take effect in a window where nothing else moved. Search Console shows you neither condition.
That is not proof that disavows do nothing algorithmically. It means a claimed recovery needs a comparison group rather than a before-and-after chart. For most sites the practical conclusion is to stop treating disavow audits as routine maintenance, keep the tool for the two situations below, and test properly if you still suspect links are the problem.
What Google’s documentation leaves room for
Google’s disavow help page ties the tool to one situation: a manual action against your site for unnatural links, or a belief that you are about to get one. It then narrows the audience: “In most cases, Google can assess which links to trust without additional guidance, so most sites will not need to use this tool.” It also warns that, used incorrectly, the tool “can potentially harm your site’s performance in Google Search results.”
Set those statements side by side and any link you might disavow sits in one of three states:
How Google treats the link now What disavowing it does Route to a recovery? Already discounted (the help page says that in most cases Google can assess which links to trust) Nothing measurable No Counted, and helping you Removes the benefit No. This is the harm the help page warns about Counted in a way that holds your rankings down Could remove the drag Possibly, but the help page does not describe this use outside manual actions Only the last row leads to a recovery, and it is the one you cannot observe. No Search Console report says a link is being held against you algorithmically. The help page also says disavowed links will continue to be shown in the Links report, so Search Console confirms that you uploaded a file, not that Google has applied it.
Since Penguin was folded into Google’s core algorithm in 2016, the working assumption among practitioners has been that unnatural links are mostly discounted rather than punished. The help page’s wording fits that assumption without proving it, which is why the last row cannot be ruled out entirely.
Why a before-and-after chart cannot carry the claim
Stripped down, a disavow recovery claim has three points on a timeline: a drop, a disavow upload, and a recovery weeks or months later. Each gap on that timeline has other candidates.
- There is no single effective date. Google says it can take a few weeks to incorporate a disavow list as it recrawls the web and reprocesses the pages it sees. The file takes effect link by link, so almost any later improvement can be claimed for it.
- Updates share the window. The last two years included several core updates, among them the March 2026 and May 2026 core updates, plus unannounced ones. Google’s core updates documentation says smaller core updates “are not announced because they aren’t widely noticeable”, and that while some changes take effect in a few days, it could take several months for Google’s systems to confirm that a site as a whole has improved. A recovery that lands in a later update fits that description at least as well as it fits the file.
- Other work ships in the same months. If the disavow went up alongside content rewrites, template fixes or pruning, the timeline cannot separate them.
- Extreme lows drift back. A site measured at its lowest point tends to look better later even if nothing it did mattered. That is regression to the mean, and it flatters any intervention made at the bottom.
- Survivorship. Recoveries get written up. Sites that disavowed and stayed down have little reason to publish.
- Lost credit is not a penalty. If rankings fell because links that used to help stopped counting, there is nothing for a disavow to lift. The file can take links out of consideration; it cannot restore value that has been discounted.
Hypothetical example:
A retailer loses a third of its non-brand clicks during the March 2026 core update. In April an agency uploads a file covering 1,400 domains flagged by a backlink tool’s toxicity score, while the in-house team rewrites 200 category descriptions. During the May 2026 core update the site regains about half the loss, and the monthly report credits the disavow. Nothing in that sequence separates the file from the May update, the rewrites or an ordinary rebound. The claim would start to hold only if the pages the disavowed domains linked to recovered while comparable pages they never linked to stayed flat.
That example also shows a quieter problem. A vendor’s toxicity score is that vendor’s model of risk, not Google’s assessment. A file built by disavowing everything above a vendor threshold risks the incorrect use the help page warns about, because it can sweep up links that were helping. A drop after that kind of file is as hard to interpret as a recovery.
Where a disavow is still a reasonable call
Outside manual action recovery, two situations justify the work, and neither is really about recovery.
- Link schemes you know about on your own profile. If you or a previous agency bought links, ran a private network or paid for guest posts with optimised anchors, those links fit Google’s link spam policy, which defines link spam as “creating links to or from a site primarily for the purpose of manipulating search rankings”. Getting them removed where you can and disavowing the rest fits the help page’s “about to get such a manual action” case. Treat it as insurance against a manual action: don’t expect a traffic change, and don’t report one as proof.
- A file that helped lift a past manual action. If a disavow file was part of a successful reconsideration request, the links it covers are presumably still out there. Deleting it to tidy up re-exposes them for no benefit.
What does not justify the work: monthly toxic-link sweeps, disavowing scraper and directory links you never built, or disavowing because rankings fell in a core update and the link profile looked like a convenient explanation.
How to test whether your file did anything
A disavow applies to the whole property, so you cannot randomise it by page. Links do point at specific URLs, though, which allows a split by target. It is not a clean experiment, but it gives you a comparison group instead of a timeline.
- Group the candidates by target. Map each suspect linking domain to the section of your site it points at. Keep only domains that link into a single section, so each
domain:line affects one side of the test. - Disavow one side first. Upload lines for the domains pointing at section A and hold back those pointing at section B. Mark the upload date with a Search Console custom annotation.
- Fix the metric before you look. Use weekly non-brand clicks and average position for the queries each section ranks for, and write down what size of gap would count as an effect.
- Set the window from the processing time. Google’s “a few weeks” is a minimum, and you cannot see when each linking page is recrawled. Read the result no earlier than six weeks after upload, which allows for that processing plus a few weekly readings, and extend the window if a core update starts inside it.
- Swap sides. If A pulls ahead of B, add B’s candidates to the file and see whether B follows after a similar lag. A second matching move is much harder to explain away than one.
Two limits apply. If the suppression you suspect works at site level, both sections move together and the test reads as null, so a null result rules out a page-level effect rather than any effect. And if both sections move at once in step with an update, the update is the better explanation.
What would change the decision:
A sustained gap in section A’s favour that opens only after a lag of weeks, does not coincide with an update, and repeats when section B is processed would justify keeping disavow work in your toolkit for algorithmic problems. Anything short of that supports treating the file as manual-action insurance and nothing more.
The position I would take
If rankings fell and the Manual actions report is empty, I would put the disavow at the bottom of the hypothesis list. Check the update calendar, template and content changes, lost links and demand first. If the link profile still explains the pattern better than anything else after that, run the split-by-target test rather than uploading a large file and waiting for a chart to agree with you. If you inherited a profile with bought links, disavow what you cannot get removed as insurance, then judge your SEO work on everything else.
Need help with this on your own site?
Niraj Raut works with businesses in Nepal, Australia, the UK, Europe and the US on technical SEO, ecommerce SEO, local SEO and AI search.
Add a reply Cancel reply
Replies are open to approved members. Log in or apply to join.