Choose the question before collecting data
Are you checking delivered placements, investigating lost references, comparing your referring domains or reviewing links to a particular page? These questions need different datasets. Define the target domain or URL, time window and source before downloading an export.
For your own site, Search Console offers a Links report. It is sampled rather than a complete live inventory, and it can include previously discovered links. A commercial crawler has its own coverage and observation times. Keep those boundaries attached to the data instead of treating either export as the entire web.
Create a traceable inventory
Keep the untouched export and work from a copy. Preserve each original source URL, destination, anchor, observed date and metric. Store missing values as missing; a blank score is not a zero score. If you combine sources, retain each observation instead of overwriting a recent finding with an older one.
| Field | Purpose |
|---|---|
| Referring page and domain | Distinguish page-level links from domain breadth. |
| Destination and anchor | Check the intended reader journey. |
| Source and observation date | Explain where the evidence came from and how old it is. |
| Current inspection | Record reachable page, detected link and any unresolved condition. |
| Decision and owner | Make the next action explicit. |
Normalize carefully and inspect important examples
Group referring domains for a breadth view, but keep source pages for the actual review. Two pages on one domain may have different content and publication terms. URL parameters, redirects and subdomains also need deliberate treatment; avoid deleting them from your raw evidence.
Open a representative set of important or unusual pages. Check the surrounding topic, link text, target and attributes. Prioritize paid placements, valuable references and recent changes. Record checks that fail because of access or timeouts as unresolved, rather than concluding that the link was removed.
Turn findings into a short action queue
Use actions that match the evidence: repair a broken destination you control, clarify an incorrect placement with its publisher, investigate an inaccessible page, preserve a useful citation, or take no action. Each task should have a URL, reason, owner and review date.
For example, an old guide on your site may redirect to an unrelated homepage. Improving that destination path may be more useful than acquiring another reference to the same broken experience. This is a hypothetical audit finding; verify the redirect before changing anything.
Do not automatically disavow links merely because a third-party score is low. A removal or disavow review has different requirements and consequences.
Repeat with a comparable baseline
Use the same scope and source settings when comparing periods. Separate genuinely new referring domains from new observations by a crawler. Mark changes in provider coverage so a jump in reported links is not presented as campaign growth.
StarBacklink My links helps you review orders made in the workspace. It is not a complete backlink-discovery index for any website. Keep that order-level record alongside your wider analysis, and use the link anatomy checklist when investigating a source page.