Backlinks in 2026: what Google and Bing still reward.

What follows is a reading of those public documents, plus two Ahrefs correlation studies, from 2025 and from 2019. It is desk research. It is not a ranking movement, a spend, or a result from a property we run. Where a figure would have to be invented — a safe anchor percentage, a share of PageRank that a nofollow link passes, a monthly quota — it is left out.

The question the spreadsheet keeps avoiding is a smaller one. Which independent pages have a reason to cite this page, and which numbers on the invoice are standing in for that reason?

Google and Bing still say links count

Google's guide to its ranking systems still describes systems that read how pages link to each other, in order to tell what a page is about and which pages might help with a query. PageRank is named there as one of the core systems, with the explicit note that how it works has changed a great deal since the original paper and that it remains in use. The same page says ranking is designed to work at page level. Site-wide signals contribute to how a page is understood. A strong site-wide signal does not make every URL on that site rank, and a weak one does not sink every URL.

Bing is plainer about reputation. Its March 2025 explanation of how results are ranked says relevance includes phrases used in links that refer to the page, not only the words on the page itself. Under quality and credibility it asks what types of other websites link to the site, and it says a well-known news site carries a higher reputation than a brand-new blog. The current Webmaster Guidelines still list external links from relevant websites as a way to be discovered, and they say strong internal and external linking supports authority evaluation. Parameters are described in a general order of importance that can change by query. That is not a weighting table.

Those two pages are enough to retire the claim that backlinks no longer matter. They are not enough to treat every backlink as an equal vote. The rest of this article is the distance between those two sentences.

Counting links is the wrong question

The easiest number to inflate is the raw backlink count. One footer link, repeated by a template, can show up as thousands of backlinks from a single site. That is one decision, copied. It is not thousands of independent recommendations. Distinct referring domains are the less gameable headline for that reason — not because a correlation study showed a dramatic gap between the two.

Ahrefs measured the association in 2025 on the top 20 results for one million high-volume US queries. These are Spearman correlations with ranking position, calculated inside Ahrefs' own index:

Ahrefs calls the figures weak to very weak, and says correlation is not causation. A coefficient of 0.255 does not mean referring domains explain a quarter of either engine. Pages that already rank are seen more often, and being seen creates links. Ahrefs also notes that well-known brands collect rankings and links together, which pulls the numbers around. The association was stronger for higher-volume queries, for local queries (about 0.33 for links and referring domains), and for informational queries than for commercial and transactional ones. Those splits are interesting. They are still one vendor's correlations, not Google's weights and not Bing's.

Read the table narrowly. Referring domains and raw backlinks are almost the same number, so the case for referring domains is the sitewide-inflation problem, not the third decimal. Domain Rating, the homepage-style score, is the weak row. Followed versus all links sit close together too, which is not a measurement of how much credit a nofollow link passes. None of the rows is a quota for how many links a page needs.

The page beats the homepage score

Domain Rating and Domain Authority describe a third-party view of a domain's backlink profile. They are a reasonable way to sort a long list of offers down to a short one. They are a poor purchasing specification, because the invoice is usually priced on the homepage score while the link lives on one URL. Google's own documentation, above, already says a site-wide signal is not the same fact as the page.

Two offers, labelled as hypothetical so they cannot be mistaken for a result we measured. Link A is a DR 85 general site that publishes a new article only to hold the placement. The URL has no links of its own and no search visibility. Link B is a DR 45 specialist site with an established guide that already covers the subject, already ranks, and already attracts links of its own. Link B can be the more useful link. Nothing Google or Bing publishes says the homepage score settles it.

Ahrefs tested a version of this in 2019, across 44,589 non-branded queries with a few thousand searches a month, looking at the top 10. URL Rating — Ahrefs' score for that URL's own link popularity — was the strongest link-related correlate in the study. Requiring the referring domain to have organic traffic did not beat simply counting referring domains, and adding the domain's traffic on top of the count changed almost nothing. That does not mean Google uses Ahrefs' URL Rating. It means stopping at the vendor's homepage metric leaves out the page the link is actually on.

Traffic still has a job, and it is a filter. A URL with sustained organic traffic is indexed and has ranked for something, which is a reason to discard empty pages. It is not a rate card. The 2019 study does not support a rule that only domains above a monthly-visits line are worth having, and neither engine has published a rule that twice the visits means twice the credit.

Relevance is the page, not the niche label

Bing's current guidelines ask for links from relevant websites, and for anchors that are relevant to the destination. Google's link documentation says the anchor and the words around it are how users and Google understand the page being cited. Neither sentence is a slogan about "niche sites" versus everyone else.

The chain that actually contains the link is the site, then the page, then the paragraph, then the anchor, then the URL being cited. A general news domain can publish a page that is precisely about the subject. A site inside the right category can publish a paragraph that has nothing to do with the destination except the anchor that was sold into it. "Same niche" is too crude a reason to buy, and "not our niche" is too crude a reason to refuse. Read the paragraph the link would sit in.

Followed, nofollow, sponsored, and ugc

There is no HTML attribute called dofollow. An ordinary link is a link with no qualifier. When the citation is a genuine editorial endorsement, that is the form we would rather have, because it is the form an engine can treat as a normal link. It is not a guarantee that the link will move a ranking.

In 2019 Google said nofollow, sponsored and ugc would be hints for ranking rather than absolute instructions, and that from March 2020 nofollow would also be a hint for crawling and indexing. The same post said that in most cases the change would not alter how nofollow links were treated, and that Google would generally not consider them for ranking. "Nofollow now passes PageRank" is a stronger claim than that post made. There is no published figure — not 10%, not 30%, not any other share — for how much a nofollow link passes.

The qualifiers still have jobs, which Google sets out in its guide to qualifying outbound links. Sponsored is for advertisements, sponsorships and other compensation. UGC is for links in comments, forum posts and other user-generated content. A nofollow link on a publication people actually read can still send visitors, and those visitors can look for the brand afterwards. That is commercial value. It is not known PageRank transfer, and it should not be entered on a spreadsheet as a fraction of a followed link.

Describe the destination. There is no anchor ratio.

Google's link documentation recommends anchors that are descriptive, concise and relevant to the page they point at, and it warns against forcing keywords into them. The surrounding sentence counts as well. Bing's ranking explanation treats phrases in the links that refer to a page as part of relevance, and its guidelines ask for relevant anchor text. Both are describing the same habit. The anchor should say what the destination is.

What neither publishes is a mix. A split of branded, partial, naked, generic and exact-match anchors, with a ceiling on the last one, is an SEO heuristic. Google's spam policies do name the manipulative version: paid posts carrying optimised anchor text, automated links, and keyword-rich links distributed widely. The working rule is the one in the documentation. Let the anchor describe the page. Do not assemble a profile so the percentages look natural.

Template links and paid ranking links

A sitewide footer or template link is one placement copied across a site. Google's spam policies list widely distributed links in footers or templates as link spam when they exist to manipulate rankings. They are the same reason a raw backlink count misleads: repetition is not a pile of separate endorsements. A citation in the body of a page that is about the subject is the placement worth wanting. A link that exists because the template repeats it is not.

Paying for attention is ordinary. Paying for ranking credit is the thing both engines name. Google's spam policies treat buying or selling links that pass ranking credit as link spam, and compensated placements are supposed to be qualified. Bing's Webmaster Guidelines say manipulating inbound links to inflate authority is a violation, and the list is link buying, link spamming, private networks, and artificial social-promotion schemes. Links, in that document, should be earned through useful content. In its December 2022 link-spam update, Google said SpamBrain could detect both the sites buying links and the sites created to pass links outward, and that the point was to neutralise those links.

A pitch built around "guaranteed dofollow, designed to move rankings" is advertising the feature that makes the placement the problem. The iGaming form of that pitch — a link inserted into an existing gambling URL and sold as authority — does not change the policy. It changes the sales sheet. Whether to buy one is a separate decision from the evidence on this page.

Rules neither engine has published

There is no official monthly quota. Bing's link-building guidance says links normally accumulate as people find the content, that a popular piece can produce a sharp increase, and that a large batch appearing at once — its example is thousands of new links in a day — is a pattern that gives the manipulation away. Google's spam policies aim at purpose and pattern: automated links, paid links, excessive exchanges, manipulation at scale. They do not publish a safe number per month. The useful question is why these links would appear, not how many the calendar is said to allow.

A link from a .gov or .edu host is not valuable because of the suffix. Universities and public bodies are often cited because of what they publish, and that reputation can be real. Bing's current documentation talks about authority, originality and the kinds of sites that link to you. It does not publish a multiplier for the TLD, and it does not publish an authority bonus that arrives when a domain turns a particular age. An old site may have years of citations and demand behind it. The registration date is not the asset.

Most toxicity-score clean-ups are a spreadsheet in search of a job. Bing removed its disavow tool in October 2023 and said it could tell natural links from unnatural ones and ignore or discount the latter. Google still provides a disavow file. The published criteria are narrow: a considerable number of artificial links, and a manual action or a real chance of one because those links violate the spam policies. A toxicity number, on its own, meets neither test. Random scraper links are part of being on the web.

The last category is the leap from an artefact to a rule. Patents about which links a reader might click, NavBoost, and leaked attributes that mention clicks are documents people can read. They do not establish that a backlink which receives more referral clicks passes more PageRank, and they do not publish a multiplier. Bing does say it looks at how people use the results page — which results they open, whether they come back, whether they rewrite the query. That is behaviour on the results page. It is not a rule about referral traffic through a backlink.

The same standard retires the folklore about how the two engines differ: that Bing weights links a fixed percent less than Google, that exact-match anchors are safe on Bing, that Bing ranks on domain age while Google ranks on relevance, that social shares are a direct Bing ranking input, or that .edu links carry a fixed bonus. Bing does publish more of the reputation language in the open, and it offers IndexNow so a new URL does not have to wait on a crawler. That discovery mechanism is the subject of large-site SEO and indexing. It is not a different link-weighting formula.

How we would evaluate a link

In this order, stopping when an earlier answer fails:

1. Would this page cite the destination if nobody were paying for a ranking? If the link exists only because of the number on the invoice, stop.
2. Is the paragraph about the subject? The site's category label is not the test. The sentence the link sits in is.
3. Is the source credible on that subject? Bing's reputation language is about which sites link to you, not about a third-party domain score.
4. Does this URL have standing of its own? Its own links, its own visibility. The homepage score is a different fact.
5. Is it a normal editorial link in the body, with an anchor that describes the destination, rather than a template or an ad that should have been qualified?
6. Is it a new referring domain, from a site that is not selling the same placement at volume? One site, repeated, is the count we already refused.

A DR floor, a DA floor, an exact-match cap, a minimum monthly-traffic figure, a links-per-month quota and a TLD bonus are not on that list. Any of them can make a report look finished while adding a rule neither engine has published.

Paid placements fail or survive a question that comes before the list. Would you still want the link if it passed no ranking credit? If the answer is no, you are buying a ranking signal. How that purchase looks in iGaming — an old URL, a contextual insertion, a permanence promise — and the checks we would run before spending, is iGaming niche edits. The search work we do sell is the other shape of the problem: whether the site can be crawled, indexed and linked from its own pages. That is SEO consultancy, not a package of placements.