Backlinks, broken down to the axes the algorithm reads.
The single backlink-count metric is a convenience. The algorithm reads the multi-axis profile: dofollow versus the four ranking-hint attributes, source topical relevance, anchor text distribution per URL, in-page link position, co-citation context, and the source's TrustRank seed-set proximity. The six axes determine whether the link passes signal, sits neutral, or actively drags ranking.
Backlinks scored on six axes, summarized in four.
A backlink is not one signal. It is a vector the algorithm decomposes across attribute type, source trust, source topical relevance, anchor distribution, link position, and co-citation neighborhood. The off-page profile that survives compounds across all six.
Dofollow, nofollow, sponsored, ugc. The four attributes that scope a link.
Every backlink carries a link attribute that scopes how Google reads it. The default is dofollow (no attribute set). The rel="nofollow" attribute Google introduced in 2005 was originally a directive to prevent PageRank flow on UGC surfaces. On 2020-03-01 Google shifted to treating nofollow, sponsored, and ugc as ranking hints rather than strict directives, meaning partial signal may still pass if the context warrants it. The rel="sponsored" attribute introduced on 2019-09-10 marks paid placements. The rel="ugc" attribute marks user-generated content. Over-using sponsored or nofollow on paid placements protects the buyer from an "Unnatural links to your site" manual action; suppressing them invites enforcement.
Domain Rating is a convenience metric; topical relevance is the signal.
Domain Rating, Domain Authority, and the equivalent vendor metrics collapse the link-graph PageRank approximation into one number. They are useful for filtering prospects but do not represent what Google scores. A DR-30 link from a topically-coherent vertical publication often outperforms a DR-60 link from a generic news aggregator. The May 2024 Content Warehouse leak surfaced mention-context scoring tied to topical relevance. Outreach campaigns get scoped against topical-cluster maps rather than raw DR target lists to maximize co-citation signals.
Anchor text breaks into six categories the algorithm scores.
The anchor-text taxonomy categorizes inbound link text into exact match (mirroring the target query), partial match (query variations), branded (the entity name), naked link (the raw URL), generic (click here, read more), and image anchors where Google uses the alt attribute as the anchor text. A natural inbound profile across a domain runs 70 percent or more branded, URL, topic, and natural-variation anchors with commercial-intent exact-match anchors at 5 to 15 percent. Penguin 4.0 evaluates this distribution per URL in real time; concentration past the per-vertical baseline triggers per-URL discounting even when the root domain looks natural.
Link velocity and decay are paired metrics.
Google evaluates link velocity to distinguish natural accumulation from manipulation. Sustained velocity from topically-relevant editorial sources reads as natural. Sudden spikes from low-quality sources trigger Penguin and SpamBrain discounting. Decay is the other half of the equation: acquired links exhibit 10 to 20 percent annual attrition as publishers edit pages, remove content, or flip dofollow to nofollow. A retainer that doesn't scope monthly placements against attrition leaks signal over time. Net-positive trajectory requires modeling placement targets against decay, not gross delivery numbers.
The natural backlink anchor baseline.
Penguin 4.0 evaluates anchor-text distribution at the URL level in real time. A landing page accumulating exact-match commercial anchors past the per-vertical baseline gets those anchors discounted the next time Google crawls them, even when the root domain looks natural. The natural distribution stays mixed across the five categories.
Methodology questions we get during the audit conversation.
What is a backlink?
A backlink is an inbound link from a third-party domain to a target site. The signal that makes backlinks ranking-relevant is PageRank flow: a link from page A to page B passes part of page A's authority to page B. Modern algorithmic evaluation layers on top: anchor text feeds Penguin distribution scoring, link-graph network position feeds SpamBrain, source topical relevance feeds the Content Warehouse mention-context scoring, and the source's own trust-graph position feeds TrustRank seed-set proximity. The single "backlink count" metric collapses these into a number; the algorithm reads the multi-axis profile.
What is the difference between dofollow and nofollow backlinks?
A dofollow link carries no rel attribute (or a rel attribute that does not include nofollow, sponsored, or ugc). It passes PageRank by default. A nofollow link carries rel="nofollow" and was originally a directive to prevent PageRank flow. Google shifted on 2020-03-01 to treating nofollow as a ranking hint rather than a strict directive, meaning partial signal may still pass if the context warrants it. Nofollow links still feed branded-mention signal and can drive referral traffic. The Google guideline requires the rel="sponsored" attribute on paid placements (2019-09-10 introduction) and rel="ugc" on user-generated content; both are also treated as ranking hints since the 2020-03-01 shift.
Are backlinks still a ranking factor?
Backlinks remain a load-bearing ranking signal across the algorithm stack. PageRank evaluates the link graph at the network level. Penguin 4.0 evaluates anchor-text distribution per URL in real time. SpamBrain reads the link-graph network to detect schemes. TrustRank scores proximity to a manually-verified seed set. The Panda patent (U.S. 8,682,892) ties backlink volume to branded query volume as a mathematical anchor. The May 2024 Content Warehouse documentation leak confirmed brand-strength scoring tied to inbound link patterns. The 2023 Helpful Content System and subsequent core updates raised the content threshold for backlinks to convert into rankings, but the off-page signal remains structurally required.
What makes a backlink high signal vs low signal?
The signal axes the algorithm reads include source domain trust (TrustRank seed-set proximity), source topical relevance (Content Warehouse mention-context scoring), anchor text and its fit to the URL's existing anchor distribution (Penguin 4.0), link position in the source page (in-content links carry more weight than footer or sidebar links), co-citation context (peer brands cited alongside the target in the source), and the source's own link-graph neighborhood. A backlink that meets all six axes carries more signal than one that meets two. DR or DA alone is a single proxy that does not capture the other five.
How many backlinks does a site need to rank?
The wrong frame. Ranking is the signal threshold against the SERP competitor set at the URL level. The relevant question is which competitor URLs rank for the target query, what their inbound profile looks like (referring domains, anchor distribution, topical-cluster proximity, branded query volume), and what closes the gap on the specific URL targeting the query. A site competing for a query where the top-10 carries 200 to 500 referring domains needs an inbound profile in that range; a site competing for a 50-referring-domain query needs less. The Panda-patent ratio adds the constraint that branded query volume must support the inbound link count mathematically.
What is link earning vs link building?
Link earning is the publisher choosing to link to a target without an outreach pitch, typically because the target site hosts a citable asset (original data, a free tool, a definitive guide) that the publisher's audience values. The link is "earned" in the sense that the asset did the work. Link building (or link acquisition) is the proactive outreach to publishers to surface assets that fit their editorial agenda. Both produce editorial placements that meet white-hat compliance. The distinction matters in scope conversations: link earning compounds when the asset surface is rich and outreach compounds when the asset surface is thin.
A backlink count is one signal. The audit reads the six.
The audit pulls every inbound link, decomposes the profile by attribute type, source topical relevance, anchor distribution per URL, and TrustRank seed-set proximity, and names the placements passing signal versus the placements that are discounted, neutral, or actively dragging.