Backlinks in 2026: what Google and Bing actually reward

The link-building industry has spent years reducing that question to numbers: DR, DA, referring domains, dofollow ratios, anchor percentages, link velocity, traffic thresholds. Some of those measures are useful. Several are proxies. Others are folklore presented with a decimal point.

Google still lists PageRank among its core ranking systems. Microsoft Bing explicitly considers the websites linking to a site when assessing reputation and uses links for relevance, authority evaluation and discovery. Neither engine, however, treats every backlink as an equal vote.

The useful question is no longer "how many backlinks do we need?" It is:

Backlinks are not dead

Google has made links less dominant than they were in the early web, but not obsolete.

Its current ranking-systems documentation still names PageRank as one of its core systems and says Google uses link relationships to understand what pages are about and which may be useful for a query. Google also describes its ranking systems as primarily page-level, while acknowledging that site-wide signals contribute to how pages are understood.

Bing is even more explicit in some respects.

Its March 2025 explanation of ranking says relevance includes the phrases used in links pointing at a page. Its quality-and-credibility assessment includes a site's reputation, asking what types of other websites link to it. Bing's current Webmaster Guidelines say external links from relevant websites help discovery and that strong internal and external linking supports authority evaluation.

So the argument that backlinks no longer matter requires ignoring the search engines themselves.

What has changed is the usefulness of simply counting them.

More links is not the same as more authority

The easiest link metric to manipulate is total backlink count.

One sitewide footer link can generate tens of thousands of backlinks from a single domain. That does not represent tens of thousands of independent recommendations.

Recent correlation research supports treating unique referring domains as the more useful headline measure. Ahrefs analysed the top 20 results for one million high-volume US queries in 2025 and reported the following Spearman correlations with ranking:

Two cautions matter.

First, these correlations are weak to modest. They do not mean backlinks explain 25.5% of Google's algorithm.

Second, correlation is not causation. Pages that rank highly receive more visibility, and visibility itself creates backlinks. Strong brands also tend to have better content, more demand and more links.

The study is useful evidence that links remain associated with rankings. It is not a reverse-engineered Google formula.

The site linking to you matters

Bing directly answers this.

Its current ranking documentation includes the reputation of websites linking to a site as part of quality and credibility. A well-known publication is not treated like a newly created unknown blog.

Its Webmaster Guidelines go further by recommending external links from relevant websites and stating that links support authority evaluation.

Google exposes less of the mechanism publicly but clearly does not treat all links as equivalent. Its spam systems are specifically designed to identify links created to manipulate rankings and neutralise whatever ranking credit those links would otherwise pass. In its December 2022 link-spam update, Google said SpamBrain could detect both sites buying links and sites created to pass outgoing links.

That immediately makes one common link-buying proposition suspect:

DR tells us something about Ahrefs' view of a domain's backlink profile.

It tells us almost nothing about:

• whether the actual linking page has authority;
• whether the page is indexed;
• whether anyone visits it;
• whether the page is relevant;
• whether the site sells hundreds of similar links;
• whether Google or Bing trusts the link;
• whether the placement has any editorial reason to exist.

A domain metric is useful for screening. It is a poor purchasing specification.

The page matters, not just the domain

This distinction is important.

Imagine two opportunities.

Link A

A DR 85 general website publishes a new article solely to place your backlink. The page has no backlinks of its own, no search visibility and no real reason to exist.

Link B

A DR 45 specialist website already has an established guide that ranks, attracts links itself and directly covers the subject of your page.

It is entirely plausible for Link B to be the more useful link.

Ahrefs tested part of this in 2019 across 44,589 SERPs. Organic traffic to the referring domain did not correlate more strongly than simply having referring domains. Page-level link popularity, represented by URL Rating, was the strongest link-related correlate in that study.

That does not prove Google uses Ahrefs UR. It obviously does not.

It suggests something more useful: the strength of the actual page containing the link should not be ignored in favour of the homepage-level metric shown by a link vendor.

Relevance matters — but be precise about what is relevant

One of Bing's clearest historical explanations came in 2012.

Microsoft used the example of a heli-skiing business receiving a link from CNN versus one from a major ski manufacturer. Bing said the CNN link could generate substantial traffic but its organic ranking value could be muted because the sites were not closely related, whereas the ski-industry link had a clearer topical relationship.

That advice is old, so it should not be treated as a specification for Bing's 2026 algorithm.

But Bing's current guidelines still explicitly recommend links from relevant websites, so the principle survives even if the implementation is unknown.

There is also an important refinement to the old example.

A general news domain can publish a highly relevant page.

If CNN publishes an authoritative investigation into heli-skiing safety and cites a heli-ski operator's original accident dataset, the page-level relationship is extremely strong even though CNN is not a skiing website.

For link evaluation, relevance is better considered at several levels:

domain → page → paragraph → anchor → destination

"Relevant niche" versus "irrelevant site" is too crude.

Dofollow versus nofollow

There is no formal HTML attribute called dofollow.

A normal link such as:

<a href="https://example.com/">Example</a>

is simply an unqualified link. SEOs call it followed or dofollow.

For direct ranking value, this remains the type of editorial link we would prefer.

Google changed the treatment of nofollow in 2019. nofollow, sponsored and ugc became hints rather than absolute instructions for ranking purposes, with nofollow becoming a hint for crawling and indexing in 2020.

That has sometimes been misreported as:

That is too strong.

Google explicitly said that in most cases the move to a hint model would not change how it treated nofollow links and that it would generally not consider them for ranking purposes.

So the defensible interpretation is:

• Normal editorial followed link: clearly eligible to contribute normal link signals.
• Nofollow: Google may choose how to use it, but direct ranking credit should not be assumed.
• Sponsored: use for advertisements, sponsorships and compensated placements.
• UGC: intended for links in user-generated content such as forum posts and comments.

There is no credible ratio saying a nofollow link is worth 10%, 30% or 50% of a followed link.

A nofollow link can still be commercially excellent. A major publication may send customers, journalists, branded searches and secondary links.

That is value.

It just should not be confused with known PageRank transfer.

Anchor text matters. Anchor-text formulas do not.

Google says anchor text helps users and Google understand the destination page. Its current link documentation recommends descriptive, concise and relevant anchors and explicitly says the words around the link matter too. It also warns against forcing keywords into anchors.

Bing likewise says relevance includes phrases used in links referring to the page.

What neither engine publishes is anything resembling:

• 50% branded;
• 20% partial match;
• 15% naked URL;
• 10% generic;
• maximum 5% exact match.

These ratios are SEO heuristics, not search-engine thresholds.

Google's spam policies prohibit manipulative practices such as paid advertorials with optimized anchor text, large-scale link schemes, automated links and widely distributed keyword-rich links.

The useful rule is simpler:

Let the anchor describe the destination naturally. Do not manufacture an anchor profile.

Sitewide links are not 10,000 independent votes

Google's spam policies explicitly mention widely distributed links in footers or templates as a form of link spam when used to manipulate rankings.

The Reasonable Surfer patent often cited in SEO research also introduced the broader idea that different links on the same document could receive different weights according to characteristics associated with their likelihood of being selected.

But this is where evidence discipline matters.

A Google patent demonstrates a method Google developed and could use. It does not prove the exact production weighting of a footer link versus an editorial link in 2026.

The practical conclusion survives without pretending to know the formula:

A genuine contextual citation inside the main content is a better link-building target than buying a template link replicated across thousands of URLs.

Do backlink clicks make the link stronger?

This is one of the more interesting unresolved questions.

Google has patents involving click probability. Its NavBoost system uses user interaction data in search ranking. The 2024 Content Warehouse documentation exposed many click-related and link-related attributes.

None of that proves:

And it certainly does not prove claims that clicked links pass "exponentially more equity."

The distinction is important:

1. Google can estimate whether certain links are more likely to be clicked.
2. Google uses search-result interaction data.
3. Google may have systems capable of evaluating link behaviour.
4. Therefore Google definitely multiplies individual backlink PageRank according to referral clicks.

Only the first parts of that chain have meaningful evidence. The final conclusion remains inference.

This is exactly where technical SEO frequently turns an interesting patent into a ranking rule.

Does traffic on the referring site matter?

Use it as a quality check, not as a formula.

The Ahrefs 2019 study is particularly useful here because it tested the idea directly.

Links from pages with organic traffic did not show dramatically stronger ranking correlation than ordinary referring-domain metrics. Organic traffic to the entire referring domain showed virtually no additional correlation over the number of referring domains.

This means:

is not an evidence-based search-engine rule.

But traffic remains useful operationally.

If a supposed DR 70 publishing site has:

• almost no ranking pages;
• no readership;
• dozens of new casino, CBD, loan and crypto articles every week;
• obvious paid outbound links;

its lack of visibility is a useful warning about the site.

Traffic is therefore a vetting signal, not known link equity.

Link velocity is mostly misunderstood

There is no official monthly backlink quota.

A website covered by international media can acquire thousands of backlinks overnight without having done anything manipulative.

Bing's own link-building documentation makes essentially this distinction: organic links normally accumulate as content is discovered, but something becoming popular can produce a dramatic increase in inbound links. At the same time, Microsoft warns that suddenly manufacturing thousands of links is an obvious spam pattern.

Google similarly targets the purpose and pattern of links rather than publishing a safe acquisition rate. Its spam policies focus on automated links, paid links, excessive exchanges and scaled manipulation.

A better question than:

is:

If there is a convincing answer, velocity itself is not the problem.

.edu, .gov and old domains are not magic

Institutional websites often make excellent sources.

Governments, universities and established organizations tend to accumulate citations because they publish information that other people trust.

That does not establish an algorithmic .gov or .edu multiplier.

Historical Bing material placed considerable emphasis on trusted institutional sources, and old SEO guidance often converted that into "Bing loves .edu links." Current Bing documentation instead talks about authority, reputation, originality and the kinds of sites linking to you.

The same applies to domain age.

An established website may have twenty years of:

• links;
• citations;
• content;
• branded demand;
• user history;
• reputation.

Those things can matter.

The fact that the domain was registered twenty years ago does not mean the registration date itself deserves the credit.

Claims that Bing gives a special authority increase at exactly three years are not supported by its current documentation.

Paid links: advertising is fine; buying ranking credit is not

Google draws a clear distinction.

Buying advertising is normal.

Buying links for ranking purposes is link spam. Paid placements should be qualified with rel="sponsored" or nofollow.

Bing's current Webmaster Guidelines are similarly direct. Manipulating inbound links to inflate authority violates its guidelines and explicitly includes:

• link buying;
• link spamming;
• private networks;
• artificial social-promotion schemes.

This creates an awkward reality for the SEO link marketplace.

A vendor selling:

is essentially advertising the characteristic that makes the placement problematic.

What about toxic backlinks?

Mostly, stop staring at the toxicity score.

Bing removed its Disavow Links feature in October 2023. Microsoft said improvements to its systems meant it could distinguish natural from unnatural links and ignore or discount the latter without affecting legitimate ones.

Google still provides a disavow mechanism, but its modern spam systems similarly aim to neutralise artificial links rather than simply treating every low-quality inbound link as negative equity. Google's 2022 link-spam update described SpamBrain as neutralising unnatural links and removing the credit they had passed.

That makes routine spreadsheet exercises such as:

difficult to justify.

A site with a history of deliberate paid link schemes or a manual action is different.

Random scraper links are part of being on the web.

Google and Bing are more similar here than SEO folklore suggests

There are differences between the engines.

Bing publicly exposes some signals Google discusses less directly. Microsoft explicitly says phrases in links contribute to relevance and that referring-site reputation contributes to quality and credibility. It also provides IndexNow, making discovery less dependent on waiting for crawlers to encounter new URLs.

Google gives us PageRank, detailed link-spam policies and the nofollow/sponsored/UGC framework.

What we cannot responsibly say is:

• Bing values backlinks 30% less than Google;
• exact-match anchors are safe on Bing;
• Google cares about relevance but Bing cares about domain age;
• social shares directly increase Bing rankings in 2026;
• Bing gives .edu links a fixed authority bonus.

The evidence is not good enough.

The overlap is much larger than the difference:

Earn links from credible sources, on pages relevant to the subject, where the citation exists because it helps the reader.

What a good backlink actually looks like

We would evaluate a prospective link roughly in this order:

Notice what is missing:

DR threshold. DA threshold. Exact-match ratio. Minimum monthly traffic. Maximum links per month. .edu bonus.

Those can make a backlink spreadsheet look sophisticated while adding very little knowledge.

The link-building strategy that survives every algorithm update

The safest way to build links is inconveniently similar to ordinary marketing.

Create something worth citing.

That can be:

• original data;
• an industry dataset;
• primary research;
• a useful tool;
• statistics maintained better than anyone else's;
• a regulatory database;
• original reporting;
• a genuinely authoritative guide.

Then put it in front of people whose work has a reason to reference it.

The resulting structure looks less like traditional "link building":

useful asset → distribution → coverage → citations → secondary citations

and more like publishing.

That distinction matters.

Google says links created primarily to manipulate rankings are spam. Bing says links should be earned organically through useful content.

The closer a link-building programme gets to paying websites because their third-party authority score is high, the further it moves from what either search engine says it wants.

The practical conclusion

Backlinks still matter in 2026.

Ahrefs' million-SERP study found referring domains and backlinks remain meaningfully associated with ranking, particularly at higher-volume queries and for local searches. Interestingly, the study found backlink correlations were higher for informational queries than for commercial and transactional ones, a useful counterpoint to the assumption that links matter most on money pages.

But counting links is now one of the least interesting parts of backlink analysis.

A better hierarchy is:

editorial legitimacy
→ topical relevance
→ source/page authority
→ contextual placement
→ referring-domain diversity
→ natural anchor text
→ link qualification
→ raw quantity

That does not mean every relevant backlink will move rankings or that every authoritative unrelated link is useless.

It means the evidence does not support reducing a link to:

The web is a graph of recommendations.

Google and Bing have spent decades trying to determine which recommendations are real.

The closer your backlinks resemble genuine citations that would exist without an SEO industry, the less you have to care which anti-spam system gets updated next.