A lost backlink is a reported change, not a complete diagnosis
A lost backlink is a link that a dataset previously observed but no longer counts in the same way. The label needs a provider, date and scope. It can describe an actual removed reference, a source page that could not be checked, a changed URL or a change in the provider’s index. Do not translate every loss label into “the publisher deleted my link.”
This guide starts from a supplied record. It helps reconcile a report with current evidence; it does not discover every link your domain ever had. Monitoring known pairs and rebuilding an unknown historical inventory are different jobs. Keep the original record even when later evidence changes your conclusion.
Keep three clocks and two identities separate
Record when the link was first observed, when the report called it lost and when you independently rechecked it. Preserve exact source and target URLs at each stage. A new source address, a changed destination or a redirected page can explain a disagreement without proving that the same reference disappeared. Record the collection method alongside the date.
Ahrefs documents several loss reasons, including source availability and index changes. These are its reporting definitions, not universal search-engine verdicts. Its visible-but-lost explanation also identifies rendering and access differences. Read the specific reason in your own export before deciding which observation to collect next.
A dated example: the report and browser can both be accurate
The following authored timeline uses a fictional reference from publisher.example/resources to resource.example/soil. All dates and observations are examples. The report and browser checked under different conditions: a failed crawler request on one day does not contradict a successful browser view on another.
The latest successful observation supports keeping the reference under monitoring. It does not prove that the report was wrong at its collection time, that the link was continuously present or that any search benefit was restored. Keep the earlier access failure visible and set a next check using a comparable method.
| Date and method | Observed evidence | What it establishes | Next step |
|---|---|---|---|
| 2026-10-20 · saved HTML | Exact source link present | Baseline for this pair | Retain source, href and capture |
| 2026-10-24 · report crawl | Source returned 403; link unknown | Access failure at collection time | Investigate the failed request |
| 2026-10-26 · browser review | Reference present; target 200 with relevant content | A later working visitor path | Repeat a comparable check; retain the earlier report |
Choose the next investigation from what was actually observed
If the source cannot be read, investigate access or availability; link presence remains unknown. If a readable source has no matching reference, compare its version and rendered page before classifying removal. If the exact href changed, record old and new destinations and inspect their purpose. If the link is present but its destination fails, use the separate broken-backlink repair workflow.
If a provider stopped counting a page while the link remains visible, investigate its definition and collection scope. A provider index change does not establish the page’s Google index status. If a rendered browser shows a link absent from server HTML, retain both observations and their methods. Do not overwrite one with the other.
Turn a loss alert into an incident someone can close
Assign a source owner, destination owner or research owner according to the problem. Define the next check and the evidence that would resolve it: a fresh readable source, a confirmed href change, a destination repair or a documented explanation of report scope. A publisher contact should describe the precise reference and context rather than demand restoration from a dashboard label.
The example CSV retains the three dates and observations; the blank version gives you the same columns for a new incident. Add your source, target, provider reference and owner in your working register. Use the existing monitor for known pairs and preserve exported evidence outside its retention window when needed. No download sends a message or performs a live check.
Download the worked incident CSV · Download a blank incident CSV
Measure observed recovery separately from ranking outcomes
A later matching reference is a new observation. Record what was found, the response, method and date; avoid claiming continuous availability between checks. Close the incident against its acceptance condition, while keeping unresolved collection differences open. One recovered reference is not evidence that a traffic decline has been explained.
Google’s Links report also has sample and historical limits. Compare sources within their own scope and dates rather than subtracting unrelated totals. The monitoring guide explains our saved-link checks; the repair guide handles failing destinations; the removal guide covers requests to change references. This page supplies the investigation between a loss label and one of those actions.