Disavow culture still hangs around SEO Slack channels like an old superstition. Someone exports ten thousand referring domains, sorts by a toxicity score, and uploads everything below an arbitrary threshold. Weeks later, rankings are unchanged or worse, and nobody can say which useful links were collateral damage.
Google has said for years that the disavow tool is for exceptional cases, mainly when you have a history of paid or otherwise manipulative links you cannot get removed. That guidance still holds in 2026. A serious backlink audit treats disavow as a last instrument, not the default deliverable.
What a backlink audit is for
A useful audit answers four questions. Who links to you at the domain level? Which anchors concentrate risk? Has link velocity or network footprint changed in a way that looks unnatural? Which competitor link gaps are real opportunities versus vanity mentions?
It also separates operational workstreams: removal outreach for links you can influence, on-site cleanup if you still operate old link schemes, documentation for a reconsideration request if a manual action exists, and only then a carefully edited disavow list.
Ignore vanity metrics; read patterns
Third-party toxicity scores can help triage, but they are not courtroom evidence. Domains get misclassified. News sites, forums, and niche blogs often look “spammy” to classifiers while remaining normal referral sources. Conversely, some clean-looking domains participate in private blog networks.
- Prefer referring domain reviews over raw backlink row counts.
- Inspect anchor text distribution for over-optimised commercial phrases.
- Check whether low-quality links share IP ranges, WHOIS patterns, or footprint templates.
- Compare your profile to competitors in the same market before calling something abnormal.
If traffic dropped and you suspect links, also test whether the drop matches a manual action, an algorithmic update, or an on-site failure. Mixing those causes leads to the wrong cleanup. Use manual action versus algorithmic suppression as the first fork in the road.
Build risk tiers before you touch disavow
| Tier | Typical signals | Default action |
|---|---|---|
| Clear manipulative | Paid footprints, PBNs, hacked widgets, sitewide commercial anchors you bought | Attempt removal; disavow if removal fails and evidence is strong |
| Suspicious cluster | Many low-quality domains with identical anchors or templated content | Sample deeply; document; selective disavow only for the cluster core |
| Low quality, likely ignored | Random directories, expired domains with odd scraps, spam comments | Monitor; usually no disavow unless part of a wider scheme you created |
| Normal or useful | Editorial mentions, partners, community sites, niche references | Keep; never bulk-disavow because a score looked scary |
Tiering forces humility. Most profiles contain junk. Search systems are built to ignore a lot of junk. Your job is to find the junk that looks like deliberate manipulation, especially manipulation your team or a past vendor created.
Anchor text and destination concentration
Anchor text still matters as a risk signal when it is unnaturally repetitive. A sudden pile of exact-match commercial anchors from unrelated domains deserves attention. Branded anchors, naked URLs, and natural phrases usually do not.
Also watch destination concentration. If thousands of low-quality links point at a single deep URL you once tried to push, that campaign history matters. If junk links point at the homepage with generic anchors, the practical risk is often lower than the export volume suggests.
A sane sampling method
- Export referring domains from at least one major link index and Search Console links data.
- Sort by domain authority proxies if you must, but re-sort by shared footprints next.
- Open 20 to 50 sample pages per suspicious cluster and record why they look manipulative.
- Tag domains as remove, disavow candidate, monitor, or keep.
- Only domains tagged disavow candidate enter the file draft.
When disavow is justified
Disavow makes sense when you have a manual action for unnatural links, or when you can document a past link-building scheme you cannot fully unwind through removals. It can also make sense for a narrow, well-evidenced cluster that continues to grow after outreach fails.
Disavow is a weak response to ordinary directory spam you never bought. It is a dangerous response when your evidence is only “the tool said toxic.” Wrong disavows can ask Google to ignore links that were neutral or helpful. You will not get a clean changelog telling you what you broke.
Example disavow file structure with comments for reviewers
# Disavow draft - 2026-06-18
# Scope: paid guest post network used by prior vendor (2023-2024)
# Evidence: shared templates, invoice trail, identical commercial anchors
# Removals attempted: 2026-05 (partial success)
domain:pbnsample-one.example
domain:pbnsample-two.example
# Individual URL only when the domain also has legitimate sections
https://mixed-site.example/sponsored/best-widgets-cheap/Keep comments in the file for humans on your team. Future you should know why each entry exists. If you cannot write that comment, the domain probably does not belong in the file yet.
Removals, documentation, and recovery paths
For links you control or can influence, removal or nofollow requests come before disavow. Keep a spreadsheet with contact dates, responses, and final status. If you file a reconsideration request, that log becomes evidence that you did substantive work.
If there is no manual action and rankings are soft after an update, link cleanup alone may miss the real cause. Content quality, scaled doorway patterns, or technical index issues can look like “bad links” from a distance. Pair link review with a penalty and recovery audit when the drop is severe and the cause is unclear.
The goal of a backlink audit is a defensible story about risk. A giant disavow list without a story is just a hope that Google will sort out your uncertainty.
Competitor gaps without copying risk
The constructive half of a backlink audit is opportunity mapping. Which domains link to multiple competitors but not to you? Which resource pages still accept relevant additions? Which digital PR angles earned links in your niche over the last year?
Chase editorial relevance, not raw domain scores. Copying a competitor’s obvious paid placements is how teams recreate the problem the audit was meant to solve. If your wider SEO plan is still deciding between a one-off cleanup and ongoing work, read SEO audit or ongoing SEO before staffing a permanent link-building machine.
Read the audit as a risk register and an opportunity list. Act with restraint on disavow. Act with energy on removals you can finish and on earned links worth pursuing. That balance is how you avoid cutting the wrong relationships while still taking manipulative history seriously.