Referring domains vs backlinks: count the right thing

A backlink is a link from another website to a target page. A referring domain is a distinct source domain contributing links to the target being measured. One domain can supply many backlinks, so the two counts answer different questions. Always check the target scope, grouping rules and capture date before comparing them.

The distinction appears in Ahrefs’ definitions and in the grouping described by Google’s Search Console Links report documentation, reviewed on 26 September 2026. The useful next step is reconciling the units in your own report.

What each count answers

Use backlink records when you need to inspect an individual placement. Use a referring-domain count when you need to know how many distinct source domains appear under a declared grouping rule. Neither count, on its own, tells you whether the links are useful, current or editorially relevant.

A fictional count reconciliation worksheet

The following invented records illustrate the arithmetic. All source names use reserved example domains. Assume each row is one distinct captured link observation, and that the two source hosts are grouped under publisher.example.

For the entire target domain, this worksheet contains five link observations, two source hosts and one grouped referring domain. For the target guide page alone, it contains three observations, two hosts and one domain. Changing the target scope changes some counts while leaving others unchanged.

Add capture date and source-dataset columns when using this worksheet with real records. If one observation came from a different crawl period, do not quietly present all five as a simultaneous live state. If your tool deduplicates links differently, document that rule before reconciling totals.

Consider another illustrative case: one source page contains the same target URL twice, once in its main text and once in its footer. An occurrence-based extraction could record two HTML links; a rule that keeps unique source-page and target-URL pairs would record one pair. Both observations still come from one source host and one grouped source domain. If a second page on that host links to the same target, the pair count becomes two under that rule, while the host and domain counts remain one. State which deduplication rule your report uses before comparing its backlink total with another tool.

Subdomains need an explicit grouping rule

Do not derive every registrable domain by taking the final two labels. Registration boundaries vary: co.uk is a familiar example. The Public Suffix List explains why suffix-aware handling is needed. Use the host-versus-domain analysis worksheet when source hosts need to be cleaned and grouped before counting. A domain grouping rule also does not establish whether the sites share a legal owner.

Common Crawl’s domain graph aggregates hosts at pay-level domain using that list. Its graph can therefore answer a different counting question from a report that preserves individual source pages. Our Common Crawl web-graph guide shows where this aggregation sits in the data pipeline.

Why two tools can disagree

Check scope before judging accuracy. One report may cover a single page while another covers a domain. One may combine subdomains; another may list hosts separately. They may use different crawls, time windows, deduplication or retention rules. A larger total is not enough to explain which method is better for your question.

Search Console illustrates the need to read the method. Google says its Links report contains a sample, groups sites by root domain, combines duplicate links and can include links that have disappeared. It is useful evidence for a verified property, but the report is not a promise of a complete current backlink inventory.

Reconcile a small set of source records first. Can you explain which target each report covers and which observations are combined? If those definitions remain unclear, label the totals as noncomparable rather than calculating a precise-looking growth percentage.

Choose the unit for the next action

For broken-link investigation, a domain count is too coarse. You need the source page, target URL and enough current evidence to determine whether that placement still exists. A report showing one referring domain could conceal several placements with different destinations and different states. Keep those records separate until the page-level task is resolved.

For comparing the breadth of observed sources, a referring-domain count is easier to interpret than a raw placement total. A navigation link repeated across many pages can generate numerous observations from the same source. Grouping those observations prevents that one source from looking like many independent websites. It still leaves source relevance and the meaning of the links to be assessed.

For investigating subdomain structure, retain host-level records too. A publication’s news and blog hosts may have different content, even when your final report groups them into one domain. You can present a grouped summary while keeping the underlying hosts in the worksheet for explanation. Declare the grouping rather than making readers guess which names disappeared.

Do not choose a universal backlinks-to-domains ratio as a target. The useful mix depends on the question and the observations being counted. A ratio calculated from mismatched scopes or capture periods can be numerically correct and still be meaningless for the decision. Report the underlying units before offering an interpretation.

Make the comparison reproducible

Before handing the figures to a colleague, write a measurement note with four parts: the target, the source, the counting rule and the date. An example template is: “Target scope: entire domain. Source: named report or graph release. Unit: distinct grouped source domains under the stated normalization. Date: capture period where available, plus export date.” Fill every field from evidence.

Then keep a small reconciliation sample with the report. Include at least one source that contributes multiple observed links and, if present, one domain containing multiple source hosts. These examples explain how your counting rule behaves. They are documentation choices, not a statistically representative quality test.

If a colleague cannot reproduce the total because the tool exposes only an aggregate, say so. You can still quote that aggregate with its provider and scope, but you should not claim an independently verified complete count. This boundary matters more than adding extra decimal places or comparing two totals without compatible definitions.

Shared linkers answer another question

A shared-linker count compares the incoming source lists of two targets. It answers how many source domains occur in both lists under the stated method. That is different from counting all referring domains to one target.

LinkCensus uses dated graph editions for related-domain discovery and shows shared linkers (among top 200 retained). This is bounded explanatory evidence. The co-citation guide explains the intersection and union calculation, including why a retained overlap count and a normalized score have different limits.

Use LinkCensus to investigate related domains and read the product method before comparing scores. For a business decision, follow the competitor validation worksheet: graph proximity does not establish that a candidate sells to the same buyers.

What are referring domains in SEO?

Referring domains are distinct source domains with links to the target page or site being measured. A single referring domain can contribute several backlinks. The count depends on the report’s target scope, domain grouping and data coverage, so include those definitions when quoting it.

For your next report, put the counting unit beside the number. “Distinct referring domains observed in this dated dataset” gives a reader something they can evaluate. “Links” alone leaves too much of the measurement unspecified.