How Tools Count Nofollow, UGC, and Sponsored Links
Almost every backlink total you see includes nofollow, ugc, and sponsored links by default. The attributes are recorded as a property of each link row, filterable after the fact, and generally not excluded from the headline count. So “2,140 backlinks” and “2,140 followed backlinks” are different claims, and the report usually shows you the first while the reader assumes the second.
That gap is the whole subject of this post.
What the attributes are, as data
Since 2019, three values are in common use in the rel attribute: nofollow, ugc for user-generated content, and sponsored for paid or advertising placements. Google documented all three at the time and stated that nofollow would be treated as a hint rather than a directive for ranking purposes.
Two consequences for measurement follow directly.
The attribute is observable, the treatment isn’t. A crawler can read rel="nofollow" from the HTML with certainty. What a search engine does with it is a hint-based decision that is not exposed anywhere. So your data has a clean field and a completely opaque meaning.
“Followed” is not a synonym for “counted.” A link with no rel attribute may still be discounted for a dozen reasons a tool cannot see. Filtering to followed links narrows the set to those without a declared hint, which is not the same as the set that carries weight.
Where the counting gets messy
Even the observable part has edge cases, and vendors handle them differently.
Multiple values. rel="ugc nofollow" is legitimate and common on forum software. Whether a tool files that under one bucket, both, or a compound category varies.
Page-level directives. A <meta name="robots" content="nofollow"> applies to every link on the page without any of them carrying the attribute. Some indexes reflect this; others record the individual links as followed because the markup on the anchor says nothing.
JavaScript-inserted attributes. If the attribute is added client-side, a crawler that doesn’t execute scripts sees a followed link. A crawler that does may see something else.
Redirected links. A link may point at a redirector that strips or adds attributes on the hop. What the tool records depends on how far it follows and what it attributes the final destination to.
Historical rows. Links indexed before 2019 predate ugc and sponsored entirely. Old rows may carry a plain nofollow classification for links that would be labelled differently if crawled today.
None of this is a vendor failing. It’s the ordinary consequence of parsing a messy web, and it’s part of why two indexes disagree even on a field that looks binary — see why two tools report different backlink counts.
The reporting trap
Here’s the sequence that causes trouble. Someone reports total backlinks and referring domains from the default view. Later, someone else applies a followed-only filter and gets a number 30% lower. The obvious conclusion — that links were lost, or that the earlier number was inflated — is wrong. Two different filters were applied to the same index.
A hypothetical illustration: 2,140 links from 812 referring domains at default settings; filtered to followed-only, 1,490 links from 604 domains. Nothing has changed except the filter, and both rows are correct statements about different sets. (Illustrative figures.)
The fix is mechanical: state the filter in the label, every time. “Referring domains (all attributes)” and “referring domains (followed only)” as separate rows, both present, both dated.
Should you filter to followed-only?
Report both, and lead with whichever answers the question.
Report all-attributes when the question is about visibility or citation. A nofollowed mention in a major publication is a real citation with real referral traffic and real brand effect. Excluding it from a “who cites us” analysis loses information for no gain.
Report followed-only when the question is about link position. This is the closer proxy for what a ranking system might weight, with the loud caveat that “no declared hint” is not “counted.”
Never report only followed-only and call it your backlink profile. You’ve silently deleted a category of citation that includes most of social, most of forums, and most of press with strict linking policies.
Never treat the attribute as a quality grade. Plenty of nofollowed links sit on excellent pages; plenty of followed links sit on junk. The attribute describes a publisher’s policy, not the page.
What the distribution can tell you
The followed/nofollowed ratio, read as a distribution rather than a target, occasionally says something.
A profile that is almost entirely followed, from unrelated domains, with commercial anchors, has a shape that ordinary citation doesn’t produce — most ordinary citation arrives with a substantial nofollowed fraction, because platforms default to it. This is a pattern-level observation of the kind discussed in anchor text distribution as a dataset, and it works the same way: uniformity is the signal, not a threshold.
There is no published ideal ratio. Anyone quoting one is quoting folklore, and the natural ratio for a consumer brand with heavy social presence and a B2B vendor cited in trade press are obviously different.
What nobody outside Google knows
How much a hint discounts a link, whether it varies by context, whether sponsored is treated more strictly than nofollow, whether the attributes affect discovery and crawling as well as ranking, and whether a followed link from a page full of paid placements carries anything at all.
Given that, the defensible position for a report is narrow: the attribute is a documented, observable property of the markup; the search engine’s use of it is a documented hint with undocumented weight; and every count you publish should say which attributes it includes. That last part is entirely under your control, and it’s the part that gets skipped.
For the broader habit this fits into, see building a link report you can defend.