What is a backlink?

A backlink is a link on another website that points to a page on yours. Search engines read it as a vote and as a description of your page, but how much it counts depends on who links, where the link sits and what it says.

In short

  • A backlink is a hyperlink on someone else's site that points to yours, and a referring domain is a separate website with at least one such link.
  • Referring domains correlate with rankings slightly better than raw backlink counts, at 0.255 against 0.248 in Ahrefs' January 2025 study of one million keywords.
  • A "dofollow" link is simply a normal link with no nofollow, sponsored or ugc label, and Google has treated those three labels as hints since 2019.
  • Google's spam policies list widely distributed footer and template links as link spam, so placement inside the main text matters.
  • Most poor links are ignored rather than punished, which means the usual cost of a bad link is the money spent on it.

A backlink is a hyperlink on another website that points to a page on yours. You will also see it called an inbound link or an inlink. To a search engine it carries two messages: the linking site is willing to send its readers to you, and the words of the link say something about what your page covers. Not every backlink counts the same, and some do not count at all. A link is a small thing, a few words on someone else’s page, yet the question of which links count is what most of link building turns on, so the vocabulary is worth learning properly before any money or effort is spent chasing them.

A referring domain is a separate website that links to you at least once. If one blog links to you 100 times, you have 100 backlinks and one referring domain. Beginners often meet both numbers in a tool on the same screen, see that one is far larger than the other, and reasonably ask which to trust. The larger number is easy to inflate without anything meaningful having happened. A single site that adds your link to its sidebar may create a backlink on every page it has, and nothing about your standing has changed except that one site owner made one decision.

The distinction matters because the first link from a site says much more than the fiftieth. In Ahrefs’ study of one million US keywords, published in January 2025, the number of referring domains correlated with rankings at 0.255, slightly ahead of the raw number of backlinks at 0.248. The gap is small, but it runs in the direction you would expect if search engines care more about how many independent sites vouch for you than about how many times they do it. Most link builders therefore count referring domains, not links. A correlation is not proof of cause. Pages that rank well also attract links because more people see them, so some of the relationship runs backwards, from ranking to links.

An exhibit from the US antitrust case against Google, reported in May 2025, shows links working in two places. The first is trust. The exhibit describes PageRank as “a single signal relating to distance from a known good source”, used as an input to a quality score for the site. PageRank is Google’s link-based scoring system, and the phrase means that a site is judged partly by how short the chain of links is between it and sources Google already regards as reliable. The second place is relevance. Anchors, the clickable words of links, are one of three core relevance signals alongside the body text of the page and user clicks.

So a backlink can raise trust in your whole site and, separately, help one page rank for the topic named in the anchor. The two effects work at different levels, and it helps to keep them apart. As an invented example, a small site sells handmade candles. A link from a respected interiors magazine to its homepage, using only the shop’s name as the anchor, adds to the trust side for every page on the site without describing any page in particular. A link from a small craft blog to its page on beeswax candles, using the words “beeswax candles”, does less for trust and more to tell Google what that one page is about. How search engines evaluate links covers this in detail.

Followed, nofollow, sponsored and ugc

A normal link has no special label. People call it a “dofollow” or followed link, although no such attribute exists in HTML, the code that web pages are written in. The word only exists as the opposite of nofollow. A site can add one of three labels to a link: nofollow (we do not vouch for this), sponsored (this was paid for) and ugc (a user added this, as in a comment). Readers never see these labels, which sit in the code of the page for search engines to read. Since 2019 Google has read all three as hints, so it decides for itself whether to use the link.

That leaves an obvious question about whether labelled links are worth having, and the data does not settle it. In Ahrefs’ 2025 study, followed referring domains correlated with rankings at 0.250 and all referring domains at 0.255. The two numbers are too close to tell whether labelled links pass value. If labelled links were plainly worthless, you might expect that counting only followed domains would give a clearer relationship with rankings, and it does not. Equally, nothing in the figures shows that labelled links help. The sensible position is to stop treating the label as the deciding factor. A natural link profile contains both kinds, and one made up entirely of followed links would itself look arranged. The full detail is in link attributes: nofollow, sponsored and ugc.

The exact weights are not public. The table sets out the features that Google’s own documents, court exhibits and the 2024 leak point to, with a note on how firm each one is. The third column is there because the evidence is uneven: some features are confirmed in Google’s own words, and others are well supported in direction without anyone knowing how much they count.

Feature What to look for How firm the evidence is
Authority of the source The linking site is itself linked to by trusted sites Confirmed. The court exhibit describes PageRank as distance from a known good source
Relevance The linking page and site cover your topic Strong in direction. Google’s guidance and several leaked fields agree, but no weighting is known
Placement The link sits in the main text, where a reader would click it Strong in direction. Google’s spam policies name footer and template links as link spam
Anchor text The words describe the target page honestly Confirmed as a relevance signal. Heavy keyword anchors are also the classic spam pattern
Editorial choice A person at the site decided to link, with no payment or contract This is the line Google’s spam policy draws
Indexed source page The linking page is in Google’s index Logical requirement. A link on a page Google has not indexed is unlikely to count
A new domain The site has not linked to you before Supported by the Ahrefs correlation above

Two features are often overrated. A high Domain Rating or Domain Authority on the linking site is a third-party estimate, not something Google uses, and it says nothing about the specific page that links to you. These are scores that SEO tools calculate from their own data, and a site with a high score can still hold thousands of pages that nobody links to or reads. See Domain Authority vs Domain Rating. The second is the label. A nofollow link from a publication your customers read can be worth more to the business than a followed link from a site nobody visits, because the first sends real readers and puts your name in front of the right audience whatever a search engine decides to do with it.

Google defines link spam as creating links “primarily for the purpose of manipulating search rankings”. Its spam policies, last updated in August 2026, give examples that include buying or selling links for ranking purposes, including sending a product in exchange for a link, excessive link exchanges, links created by automated programs, low-quality directory or bookmark site links, forum comments with optimised links in the post or signature, and widely distributed links in footers or templates. What ties the examples together is that in each case the link exists because someone wanted a ranking benefit, and no reader was meant to use it. Buying a link for advertising is allowed when it is labelled sponsored or nofollow, so the policy objects to hidden payment, not to payment as such.

What happens to links on the wrong side of that line is less dramatic than many people assume. Since the Penguin update became part of the core algorithm, Google has mostly ignored such links instead of punishing the site. A manual action, applied by a human reviewer and shown in Search Console, still exists for profiles dominated by manipulative links. Google link penalties and recovery explains both. For most sites, then, the usual cost of a bad link is the money spent on it. The link sits in your reports looking like any other and contributes nothing, which is a poor outcome but a survivable one.

The idea that every weak link is “toxic” and must be disavowed is a different matter. To disavow a link is to file a request with Google, through a tool it provides, asking it to disregard that link. Google’s John Mueller has called toxic links a concept made up by SEO tools, and Google’s guidance limits the disavow tool to sites with a manual action or a large number of links they built or paid for themselves. Every site that has been online for a while collects odd links from scrapers and low-grade directories without doing anything to invite them, and if Google mostly ignores poor links there is little to gain from listing them. A beginner who sees a tool flag hundreds of “toxic” links should read that as the tool’s opinion, with no penalty implied.

Google Search Console has a Links report. It is free and comes from Google itself, but it shows a sample, not the full list. Ahrefs, Semrush and Majestic each crawl the web with their own bots and report what they find. They usually show more than Search Console, but none of them sees everything. Their numbers differ from each other and from Google’s because each index is different, so it is normal for one tool to report a link another has missed, and the totals should be read as estimates. Following one tool consistently tells you more than switching between them, since the trend within a single index is real even when the absolute number is not.

When you review a link, open the page it sits on. A two-minute look at the page tells you more than any score. You are checking whether the page is indexed, whether the link is in the main text, and whether the site looks like something a real audience reads. Those three checks line up with the table above, covering the indexed source page, placement and editorial choice. From here, read link attributes for the labels, then anchor text for the words of the link. If you are about to pay for a placement, start with how to vet a publisher, which turns the quick look described here into a fuller check.

Common questions

What is the difference between a backlink and a referring domain?

A backlink is one link. A referring domain is one website that links to you, however many times. One hundred links from the same site count as one referring domain.

Are all backlinks good for SEO?

No. Google says it neutralises links it considers manipulative, which means they are ignored. A profile dominated by bought or automated links can also lead to a manual action.

What is a dofollow backlink?

It is an informal name for an ordinary link that carries no nofollow, sponsored or ugc attribute. There is no "dofollow" attribute in HTML.

How can I see who links to my site?

The Links report in Google Search Console shows a sample of the links Google knows about. Third-party tools such as Ahrefs, Semrush and Majestic run their own crawlers and usually show more, but none of them sees everything.

Vendors to look at

  • BazoomEditor's pick

    Sponsored content and link marketplace, managed service

  • MotherlinkEditor's pick

    Backlink services, guest posts, niche edits, full SEO