A backlink audit that does not rely on one score

No backlink provider has a complete, real-time record of the web. Databases discover and recrawl links at different times, while Google’s own Links report is also a sample rather than a complete list. Begin by stating the source, export date, property and scope. Compare providers only as evidence sources with different coverage, not as competing counts that must reconcile exactly.

A useful audit asks practical questions: which domains link to important pages, which links have appeared or disappeared, which anchors describe the destination, and are there patterns worth verifying? It does not attempt to calculate a universal authority value or declare a site safe from a proprietary score. Link relevance and editorial context matter alongside observable technical facts such as whether the source page and destination resolve.

• Record source, export date and domain scope.
• Separate referring domains from total backlink URLs.
• Review important destination pages, not only domain-level totals.

A gained-link report can surface coverage, partnerships, citations or unexpected mentions. Open a sample of high-value sources and confirm that the link is present, crawlable and points to the intended destination. A newly discovered link may be old on the source page, and a provider may report it late. Treat discovery date and publication date as different fields whenever that distinction affects the conclusion.

A lost link needs context before action. The source page may have moved, the destination may redirect, the page may no longer exist, or the tool may simply not have seen the link recently. Prioritise losses from relevant, accessible sources that pointed to strategic pages. If a migration caused broken inbound destinations, restore the appropriate redirect. If an editorial reference was removed, consider a polite factual outreach only when there is a legitimate relationship or correction to offer.

• Verify source URL, destination URL, HTTP response and link attribute.
• Check whether the loss is a tool recrawl change or an actual removal.
• Fix your own broken destinations before requesting a new link.

Anchor text provides one clue about how a source describes the destination. Group anchors into brand, URL, descriptive phrase, generic wording and other meaningful categories, then inspect outliers in context. A concentrated commercial phrase is not automatically manipulation, and a large number of generic anchors is not automatically weak. Navigation, syndicated content, language and a site’s natural publishing habits all affect the distribution.

Look especially at the relationship between anchor, source-page topic and target-page purpose. A relevant editorial sentence pointing to a useful resource is qualitatively different from a site-wide footer link, even if an external metric rates the domains similarly. Track changes over time and preserve examples in the audit. That makes it possible to explain a concern without exaggerating what a chart can prove.

• Classify anchors consistently before comparing periods.
• Inspect source-page placement and surrounding copy.
• Use examples, not only percentages, in stakeholder reporting.

Third-party authority, spam or toxicity metrics are screening aids, not Google verdicts. They may help identify a large set for review, but a single score cannot establish that a link violates Google’s policies or causes a ranking problem. Check whether the page exists, whether it appears maintained, whether the link is editorially understandable, whether it is marked with an appropriate rel attribute, and whether the target is relevant to the page’s subject.

Patterns deserve more attention than isolated odd links: automated pages with irrelevant outbound links, paid-link arrangements intended to pass ranking credit, hacked content or repeated manipulative anchors can be reasons to investigate. Preserve URLs, screenshots where permitted, dates and rationale. Do not contact, remove or disavow links merely to improve a vendor metric. The audit should reduce uncertainty and escalate only evidence that supports a considered action.

• A metric selects a review sample; it does not decide the case.
• Review source quality, placement, relevance and pattern together.
• Keep an evidence log for any escalation.

Most unusual links require no action. Google says it can assess links and that the disavow tool should be used only when there is a considerable number of spammy, artificial or low-quality links and the links have caused, or are likely to cause, a manual action. If a manual action exists, follow Google’s documentation, address the underlying conduct and retain a clear record of outreach or corrections. A disavow file is not routine housekeeping.

For an agency or in-house team, the deliverable is a decision log: verified opportunity, monitor, fix destination, pursue editorial correction, escalate for policy review, or no action. A configurable white-label report can communicate this clearly to a client, while the underlying evidence should remain available. Repeat snapshots can show gained and lost changes, but the useful outcome is a safer, traceable process rather than a claim that every link has been judged.

• Use outreach for factual corrections or genuine relationships, not pressure.
• Treat disavow as an exceptional, documented escalation.
• Recheck redirects and target availability after migrations.