Backlinks 101

A backlink is a hyperlink from one domain to another, but treating it as a binary “got one or didn’t” misses 90% of how ranking actually works. Quality, context, anchor distribution, freshness, and the source’s own trust profile each multiply or zero out the value. Two links to the same page can differ in impact by three orders of magnitude.

TL;DR

• Quality is a function of source authority, contextual relevance, and placement — not domain age or DR alone. A link from a 200-word blog post by a topical expert outperforms a link in the footer of a DR 80 directory.
• Anchor text ratios are a survival metric. Healthy profiles trend toward 40-60% branded, 20-30% naked URL or generic, 10-20% partial-match, and only 1-5% exact-match commercial. Cross 10% exact-match and you are flagged.
• Link velocity and diversity are the second-order signals. Sudden spikes from one source type or country trigger Penguin-era spam classifiers — now baked into core ranking — within 60-90 days.

The mental model

Backlinks are like academic citations. A peer-reviewed paper cited by 200 working researchers in the same field carries weight even if it has fewer total citations than a popular Twitter thread. Google’s original PageRank paper modeled exactly this: citation networks, where a vote from an authoritative node propagates more equity than a vote from a fringe one.

Two refinements have been bolted on since 1998:

The reasonable surfer model (patented 2010, still active) weights links by where on the page they appear, the surrounding text, and the click probability. A footer link earns roughly 5-15% of the equity an in-content link does.

Topical PageRank segments authority by subject. A link from nytimes.com/cooking/ to a cookware site flows more equity than a link from nytimes.com/sports/ to the same target, because Google computes per-topic PageRank vectors, not one global score. This is why a niche-relevant DR 30 site routinely outperforms a generic DR 70 directory.

Deep dive: the 2026 reality

Google does not publish PageRank scores anymore — the toolbar metric was killed in 2016 — but PageRank still runs internally. Third-party metrics approximate it:

Metric	Vendor	What it measures
Domain Rating (DR)	Ahrefs	Logarithmic 0-100 of backlink profile strength
Domain Authority (DA)	Moz	Predictive score for ranking potential
Authority Score (AS)	Semrush	Composite of links, traffic, and trust
Trust Flow / Citation Flow	Majestic	Trust vs. raw link volume (lower TF means spam-leaning)
URL Rating (UR)	Ahrefs	Page-level equivalent of DR

None of these are the score Google uses. They are useful for relative comparisons within a single tool and useless for absolute claims. A “DR 60” link does not equal a fixed amount of ranking value.

Link attribute taxonomy (current as of 2026):

rel value	Use case	PageRank flow	Treated as ranking signal
dofollow (default, no rel)	Editorial endorsement	Yes	Yes
rel="nofollow"	Untrusted link, comments	Hint, not directive since 2019	Sometimes (Google decides)
rel="sponsored"	Paid placement, affiliate	Hint	Sometimes
rel="ugc"	User-generated content (forums, comments)	Hint	Sometimes

The 2019 change matters: Google reserves the right to follow and pass equity through any of these. They are signals, not commands. This breaks the old “nofollow = worthless” assumption — a high-quality nofollow link from Wikipedia or Forbes still influences how Google understands your entity.

Anchor text distribution that survives manual review:

Anchor type	Healthy ratio	Example
Branded	40-60%	“Acme”, “AcmeApp”, “acme.com”
Naked URL	15-25%	“https://acme.com”
Generic	10-20%	“click here”, “this article”, “their site”
Partial-match	5-15%	“Acme’s analytics tool”
Exact-match commercial	1-5%	“best CRM software”
LSI / topical	5-10%	“customer relationship platforms”

Cross 10% exact-match commercial and you trip the over-optimization filter that has been in production since Penguin 4.0 (September 2016) and is now part of core ranking. The penalty is rarely a manual action — it is a quiet demotion you will spot only by watching ranking distribution shift toward long-tail.

Link velocity matters too. Acquiring 20 referring domains a week consistently looks natural for a growing site. Acquiring 200 in one week and then zero for a month looks like a campaign. Google’s SpamBrain classifier — confirmed in the 2022 link spam update — looks for these velocity anomalies alongside footprint patterns (same C-class IP, same WHOIS, same template).

Link diversity has three axes worth tracking: source type (editorial, directory, forum, social, .edu, .gov), TLD/country, and IP class. A profile heavy on one axis is a footprint.

Visualizing it

flowchart TD
  A[Linking page] --> B{Link evaluation}
  B --> C[Source authority + topical relevance]
  B --> D[Anchor text + surrounding context]
  B --> E[Position on page]
  B --> F[rel attribute]
  C --> G[Topical PageRank vector]
  D --> H[Reasonable surfer click probability]
  E --> H
  F --> I[Equity flow decision]
  G --> J[Equity passed to target]
  H --> J
  I --> J
  J --> K[Aggregated into target page rank signals]

Bad vs. expert

The bad approach

A team buys a “DR 70 guest post package” for $300. The post lives on a multi-niche site with 400 outbound links per month, all to unrelated commercial targets. The link uses the anchor best project management software pointing at the homepage. The HTML looks like:

<a href="https://example.com" title="best project management software">best project management software</a>

Three weeks later they buy six more, all with similar exact-match anchors. By month two their exact-match ratio is 22% and their referring-domain velocity is a perfect step function.

This fails because the source is a known footprint network, the anchor distribution is now algorithmically obvious, and Google’s classifier discounts the entire batch. Worse, the discount applies retroactively — equity stops flowing from earlier purchased links too.

The expert approach

The expert earns one link per month from a topically relevant publication where the editorial decision was made by a human, not a placement broker. The anchor varies. The HTML looks like:

<p>For teams scaling past 50 people, tools like
<a href="https://example.com/enterprise">Acme</a>
offer SSO, SCIM, and audit log exports out of the box.</p>

The anchor is branded. The surrounding context is topical. The link is in-body, not footer. The source page itself ranks for relevant queries and gets organic traffic. The expert tracks the anchor distribution monthly:

// Pull anchors from Ahrefs API and compute distribution
async function anchorDistribution(target) {
  const res = await fetch(
    `https://api.ahrefs.com/v3/site-explorer/anchors?target=${target}&limit=500`,
    { headers: { Authorization: `Bearer ${process.env.AHREFS_TOKEN}` } }
  );
  const { anchors } = await res.json();
  const total = anchors.reduce((s, a) => s + a.refdomains, 0);
  return anchors.map(a => ({
    anchor: a.anchor,
    pct: ((a.refdomains / total) * 100).toFixed(2)
  }));
}

This works because every variable Google evaluates — source quality, topical match, anchor naturalness, velocity — checks out individually and as a portfolio.

Do this today

1. Open Ahrefs > Site Explorer > your domain > Backlinks > Group similar links: “One link per domain”. Sort by Domain Rating descending. The top 50 links should be the bulk of your equity. If they are not, you have a long-tail-only profile.
2. Go to Backlink profile > Anchors and screenshot the percentages. Map them to the table above. Any exact-match commercial anchor over 5% gets flagged for diversification.
3. In the same view, set the date filter to “Last 30 days” and check Referring domains velocity. If the chart looks like a single spike, you have a velocity problem — slow or pause active campaigns.
4. Open Google Search Console > Links > Top linking sites. Cross-reference with Ahrefs. Domains in GSC but missing from Ahrefs are usually genuine editorial wins worth amplifying.
5. Use Majestic to check Trust Flow vs Citation Flow ratio. A TF/CF ratio below 0.4 across many incoming links signals spam-leaning sources — investigate before they become a disavow candidate.
6. Pull a sample of 10 backlinks and verify each: is the link in body content, what is the anchor, what is the page topic, what is the rel attribute. Save the audit as a Google Sheet template to repeat monthly.
7. Set up an Ahrefs Alert (Alerts > New Backlinks) for your domain so any new referring domain pings you within 24 hours. Catch toxic patterns the day they appear, not the quarter.
8. In Google Search Console > Performance, compare branded query volume month-over-month. Backlinks that drive real traffic and brand search are the only ones that compound.