Use this guide to compare Search Console with crawler exports, investigate valuable links landing on missing pages, and review a former agency’s work. Then decide whether you need a focused recovery task, a documented disavow review, or a broader technical SEO audit.
Establish whether there is a link problem
Start with Search Console’s Manual Actions report and the site’s link-building history, not a vendor’s red warning label. A third-party toxicity score is a review cue, not a Google penalty diagnosis. Google’s disavow guidance sets two conditions: a considerable number of spammy, artificial, or low-quality links, and a manual action caused by those links or a likely manual action.
A scraper copying your article, a foreign-language directory, or a spike in unfamiliar referring domains does not establish that threshold. Google already ignores many spam links; an export does not show which links it counts. If rankings fell, compare affected pages and dates with indexing problems, redirects, site changes, and manual-action notices before blaming backlinks.
Use Google’s spam policies to investigate paid links passing ranking credit, automated link creation, excessive exchanges, and other documented schemes. Look for contracts, placement records, and repeated patterns, not merely ugly websites. The SEO reviewer owns the risk decision; a tool operator should not upload a bulk disavow file.
Reconcile Search Console with crawler evidence
Export Search Console’s external-link samples and your crawler’s backlink report with collection dates. Keep separate columns for source URL, linked target, final target, anchor, rel attributes, first/last seen dates where available, live status, and data source. Search Console and Ahrefs or another crawler cover different samples and crawl schedules; neither export is a complete inventory.
Keep original URLs and normalize host casing for matching, but do not automatically collapse HTTP/HTTPS, www/non-www, case-sensitive paths, or query parameters. Compare exact source-target pairs first, then inspect redirects. An export pointing to http://site.example/service is not yet evidence of the same destination as https://site.example/services/; verify the actual redirect path and final response.
Manually open conflicting or high-priority sources and confirm the link remains in the rendered page, where it points, and whether it carries sponsored, nofollow, or ugc. A crawler’s missing row is not proof Google rejected a link. Require this evidence trail in your SEO audit deliverables rather than accepting a raw domain list.
Recover valuable links to broken pages
Prioritize broken targets by the relevance and legitimacy of live referring pages, not a domain-score threshold. Group referring domains by target to find missing pages with multiple credible citations. A trade publication, supplier resource, or professional association may deserve attention without a flashy metric. Confirm the source still links, the destination returns 404 or 410, and the content has not moved elsewhere.
Sending every broken URL to the homepage is not backlink recovery. Restore useful missing content when appropriate, or use a permanent 301 redirect to a genuinely equivalent replacement. If no replacement serves the same need, keep the honest 404 or 410. Ask the publisher to update the link directly when practical.
The SEO reviewer and editor approve the old-to-new content match; the developer implements it. Test the exact linked URL, every redirect hop, and the final 200 response, indexability, and page content. Update internal links and sitemaps too. Repeated failures across templates or migrations belong in a technical SEO review, not a blanket disavow project.
Audit agency link building before the handover
Before an agency handover, request placement-level records: live source URLs, target URLs, anchors, dates, payments or exchange terms, outreach contacts, and renewal obligations. Ask who controls each site and whether links disappear when the contract ends. Separate links the agency created from pre-existing links; a current export alone cannot establish who commissioned a placement.
Sample suspicious campaigns at page level. Check whether editorial-looking placements were actually purchased, whether paid links have suitable rel attributes, and whether repeated exact-match anchors come from coordinated sites. Neither an anchor pattern nor a single strange domain proves a scheme on its own. Reconcile the sample with invoices, emails, and the agency’s explanations.
Before switching SEO providers, preserve all historical disavow files, removal requests, manual-action correspondence, and access ownership. Flag unexplained domain-wide exclusions for review rather than deleting them automatically. The outgoing agency supplies records; the incoming SEO reviewer documents recommendations; the business approves consequential changes. Freeze bulk cleanup until that evidence and sign-off exist.
Implement approved actions with a clear audit trail
Turn verified findings into separate decisions: retain, reclaim, request removal or link qualification, and consider disavow. For manipulative placements, seek removal or appropriate sponsored/nofollow qualification first and retain outreach records. Only approve remaining disavow candidates when Google’s threshold is met. Do not use a ranking drop as the approval criterion.
Before uploading, download the existing file, record its property, and merge approved changes into one complete UTF-8 .txt file. A new upload replaces the old list; it does not append entries. Use one URL or justified domain: entry per line, and select the correct verified URL-prefix property. The tool does not support Domain properties.
The agreed SEO audit scope should include a source-to-target evidence sheet, prioritized fixes, an owner for each task, and a post-change validation log. The developer verifies redirects and responses; the SEO reviewer checks risk decisions and manual-action status. Where a manual action exists, document cleanup for a reconsideration request. Track recrawling and outcomes rather than promising a recovery date.
Pre-implementation verification checklist
- Preserve the existing disavow file and record the relevant property, manual-action status, and review date.
- Verify the live source URL, link attributes, target response, and final destination for every proposed action.
- Approve redirect mappings only when the old and replacement pages serve the same need.
- Assign an SEO approver and developer owner, preserve supporting evidence, and define post-change checks.
Frequently asked questions
Should I disavow every link with a high toxicity score?
No. A toxicity score is a third-party model, not a Google finding. Use it to select links for inspection, then check the live page, placement history, link attributes, and campaign context. A credible audit explains why each proposed exclusion meets the risk threshold instead of treating a score cutoff as a cleanup rule.
Do I need a disavow file without a manual action?
It can be justified if a considerable number of documented manipulative links make a manual action likely. That is different from merely having scraper links or seeing a ranking drop. Review purchase records and link patterns first, seek removal or qualification where possible, and document why the remaining links warrant disavowal.
Will fixing broken backlinks restore my rankings?
Not automatically. Repairing a relevant broken target can restore a usable route for visitors and search crawlers, but the original link may not have been contributing ranking value. Verify the source, choose a genuinely equivalent destination, and monitor the repaired URL. Do not redirect unrelated expired pages simply to capture old links.
What if my previous agency already uploaded a disavow file?
Download and preserve it before proposing changes. Check the property, entries, dates, and original rationale against available agency records. A replacement upload overwrites the complete list, so merge approved changes rather than uploading a partial update. Remove old exclusions only after reviewing their evidence, not because a different tool assigns a lower risk score.