Link Signals

Why Two Tools Report Different Backlink Counts

You run the same domain through two backlink tools on the same afternoon and get numbers that differ by a factor of four. Then Search Console shows a third figure that matches neither.

The instinct is to decide which tool is right. That’s the wrong question — there’s no count to be right about. Each number is the answer to a slightly different question, asked of a different dataset, under different counting rules. Once you can name the five sources of divergence you can usually predict which tool will report more, and stop treating the gap as an error.

1. Different crawls of different sizes

No vendor has the web. Each operates its own crawler, and the size, refresh rate and priority order of those crawls differ substantially.

A link only exists in a tool’s dataset if that tool’s crawler fetched the page containing it. A page on a small, rarely-linked site may go months between crawls by one vendor and never be visited by another. Neither is malfunctioning; they’ve allocated finite crawl budget differently.

This is the single biggest contributor to raw count differences, and it’s why the gap is usually widest for sites whose links come from the long tail — forums, small blogs, regional directories — and narrowest for sites linked mainly from well-known publications everyone crawls.

Links disappear constantly. Pages get deleted, redesigns drop footers, articles get rewritten.

A tool can’t know a link is gone until it recrawls the page. So every index contains some proportion of links that no longer exist, and each vendor makes its own policy call about how long to keep a link after last confirming it, and whether to show it in a “lost” bucket or drop it silently.

Practical effect: a tool that recrawls aggressively will report fewer links than one that’s more patient, on the same site, with neither being wrong. It also means the “new links” and “lost links” reports are partly measuring the crawler’s schedule rather than the world.

This is the one most likely to produce a 4× gap.

Suppose a 9,000-page site puts your link in its global footer. Is that one backlink or 9,000?

Different tools answer differently, and some answer differently in different reports within the same tool. Options in the wild include counting every instance, counting one per source domain, counting one per unique target URL, or collapsing them with a “sitewide” flag.

If one tool tells you 1,400 and another 5,900, the first thing to check is whether a handful of source domains account for most of the difference. Sort by referring domain and look at the top rows. Very often the entire gap is two blogroll installations and a footer.

Which is precisely why referring domains is the more stable number — it’s immune to this whole category of disagreement. See referring domains versus total backlinks.

Even the definition varies:

  • Nofollow, sponsored, UGC. Included in the count, excluded, or bucketed separately.
  • Redirected links. A link to an old URL that 301s to your current one — does that count as a link to you? Tools differ.
  • Image links, JavaScript-inserted links, links inside iframes. Depends on how much rendering the crawler does.
  • Canonicalisation. Whether example.com/page, example.com/page/, and www.example.com/page are one target or three.

None of these is a judgement call with an obviously correct answer, which is why vendors made different ones.

5. Search Console is a different kind of thing entirely

Search Console isn’t a competing crawler. It’s Google reporting a sample of what Google knows about links to your site.

Three things follow:

  • It’s the only source reporting from the index that actually ranks you. That’s a genuine advantage no third-party tool can match.
  • It’s sampled and capped. The links report shows a limited set, not an exhaustive list, and the top-linking-sites view is aggregated. Don’t read the totals as complete.
  • It’s your site only. No competitor analysis. Which is most of why third-party tools exist.

So the correct mental slot for Search Console is not “a third opinion on the count” — it’s “the only view of the index that matters, at low resolution.”

What to do with all this

Pick one tool as your system of record and stay in it. Every trend line, every target, every month-over-month comparison uses that one. Switching tools mid-year produces a step change in your charts that isn’t a step change in your site.

Never compare an absolute count across tools. Compare within a tool: this month vs last, us vs three competitors, pulled the same day.

Use a second tool for discovery, not for counting. Because the crawls differ, the second tool will surface links the first missed — genuinely useful for competitor research and for finding a lost link. Just don’t add the two numbers together, and don’t average them.

When someone asks why the numbers differ, answer mechanically. “Different crawlers reach different pages, and they count a sitewide footer link differently” is a sentence that ends the conversation productively. “One of them must be wrong” starts a pointless investigation.

Check whether the gap is concentrated. A hypothetical worked example: a site shows 1,400 links in tool A and 5,900 in tool B. Sorting tool B’s export by referring domain shows two domains contributing about 4,200 rows between them — a footer link on a large directory and a blogroll on a multi-thousand-page blog. Referring domains: 206 in tool A, 214 in tool B. The profiles are nearly identical; only the counting rule differed. (Illustrative figures.)

That last check takes two minutes and resolves most of these disputes, because in most of them the disagreement was never about which links exist. It was about how to count the same ones.

toolsdata-quality