Reading the Search Console Links Report

The Links report in Search Console is the only link data that comes from Google, which makes people treat it as the authoritative count. It isn’t a count — it’s a capped, aggregated, undated summary of a subset of what Google knows, presented for diagnostic purposes rather than reporting.

Understanding four specific quirks turns it from a source of arguments into the useful cross-check it actually is.

Quirk one: the tables are capped

Each table in the report shows a limited number of rows, and the export is capped too — 1,000 rows per table, as of this writing. If you have 40,000 referring domains, you are looking at the top slice by whatever ordering Google applies, not at your profile.

This has a specific practical consequence: you cannot compute a referring-domain total from this report. The top-linking-sites table has a count of linking pages per site, and it stops. Any total you derive is a total of the sample.

It also means the report is systematically bad at exactly the thing people want it for — finding unfamiliar or suspicious domains, which by definition sit in the long tail below the cap.

Quirk two: there are no dates

The report shows what Google currently knows about. It does not show when a link was discovered, when it was last seen, or whether it still exists. There is no time series, no “new links this month,” no lost-links view.

So the natural report question — “did we gain links this quarter?” — cannot be answered here at all. You can screenshot the report monthly and diff the screenshots, which is a real if inelegant technique, and it’s the only version of trend data this source offers.

Third-party tools do carry discovery and last-seen timestamps, with their own caveats about what those timestamps mean — see what first seen and last seen actually mean.

Quirk three: aggregation hides the link itself

The top-linking-sites table aggregates by site and gives you a count of linking pages. Drilling in gives you linking pages, and drilling further gives you target pages. What you do not readily get is a flat list of link rows — source URL, target URL, anchor text — the way a third-party export gives you.

Anchor text does appear in its own table, aggregated across the profile. That aggregation is genuinely useful for spotting a distribution that looks manufactured, and useless for tracing a specific anchor to a specific link. Anchor data as a dataset is worth its own treatment: anchor text distribution as a dataset.

Quirk four: it doesn’t tell you what Google counted

This is the important one, and it’s easy to miss because the data’s provenance is so good.

The report is a list of links Google is aware of. It is not a list of links contributing to how you rank. Google has said for years that many links are discounted, ignored, or treated as hints rather than endorsements, and none of that is exposed here. Nofollow, ugc and sponsored links are not distinguished in the report; a paid link Google is confidently ignoring looks identical to an editorial one it values.

So the report answers “does Google know this link exists,” which is worth knowing and is a much weaker claim than most people extract from it.

What it’s genuinely good for

Four jobs, all of them cross-checks rather than measurements.

Confirming discovery. A link you care about — a placement, a mention, a reclaimed link — either appears in the linking-pages table for that site or doesn’t. That’s a real, first-party yes/no that no third-party crawl can give you.

Sanity-checking a third-party index. If a tool reports a top referring domain that Search Console has never heard of, one of two things is true: the tool’s crawl saw something Google’s didn’t, or the link no longer exists and the tool’s index is stale. Both are worth knowing. The general mechanism is in why two tools report different backlink counts.

Spotting the obviously anomalous. A sudden appearance of a single domain with tens of thousands of linking pages is visible even through the cap, and is usually either a widget, a template, or a scraper.

Anchor-distribution smell test. The aggregated anchor table is a cheap look at whether your profile’s anchor mix looks like something that happened or something that was ordered.

What it is not good for

• Totals. Capped.
• Trends. Undated.
• Completeness audits. The long tail is below the cap, and the long tail is where the odd stuff lives.
• Proving a link is helping. Not exposed, not knowable from here.
• Cross-tool arithmetic. Subtracting a Search Console number from an Ahrefs number produces a difference between two samples of two crawls, which is not a quantity of missing links.

How to put it in a report

Cite it as what it is: a first-party discovery check with a stated cap.

A workable phrasing, hypothetically: “Search Console lists 812 linking sites (report is capped at 1,000 rows, so treat as a floor, not a total). The three placements from this quarter’s campaign all appear, confirming discovery. Third-party index reports 2,140 referring domains; the difference is crawl coverage plus stale rows, not missing links.” (Illustrative figures.)

That paragraph is honest, uses each source for what it can support, and pre-empts the “but the other tool says” question that otherwise consumes the meeting.

The one thing to stop doing

Stop treating the Search Console number as the true value and third-party numbers as approximations of it. Both are samples. Google’s sample has better provenance and worse coverage of the tail; a vendor’s has worse provenance and often more rows. Neither is the population, and the population — links Google counts, weighted the way Google weights them — is not visible to anyone outside Google.

That’s the recurring shape of this whole topic. The data you can get describes the link graph. The thing you want to know describes a ranking system, and no report has both.