What are backlinks in SEO? How Google reads links, anchor text and nofollow

What a backlink is

Google's own definition is on the help page for Search Console's Links report: a backlink is a link on a page of another site that points to a page on your site, so it links back to you. The word is older than most SEO vocabulary. The PageRank patent that Stanford filed in January 1998 already calls the pages that link to a document its backlinks, and the links a page makes to other pages its forward links.

In HTML, a backlink is an ordinary anchor element on someone else's page:

<p>
  The clearest walkthrough we found is
  <a href="https://www.example.com/guides/tls-renewal/"
    >this guide to renewing TLS certificates</a
  >.
</p>

Every backlink has the same parts:

• The linking page, sometimes called the referring page, and the site it belongs to, the referring domain.
• The target URL in the href attribute: the page on your site the link points to.
• The anchor text, the visible words between the tags.
• An optional rel attribute, which can qualify the link as paid, user-generated or not vouched for.

SEO tools count backlinks two ways: every linking page, and referring domains, the number of distinct sites that link to you. A hundred links in one site's footer are a hundred backlinks and one referring domain. Search Console's Links report counts sites by root domain, so links from www.example.com and m.example.com are both listed under example.com.

For Google to use a backlink at all, it has to be able to crawl it. Google's link best practices say it can generally crawl a link only if it is an <a> element with an href attribute that resolves to a real URL. A <span> with a click handler, or an <a> whose destination exists only in a JavaScript onclick, may never be extracted. Links inserted with JavaScript are fine as long as they end up in that markup. The linking page matters too: a page whose robots meta tag carries nofollow (or the same rule in an X-Robots-Tag HTTP header) asks Google not to follow any of its links.

Backlinks, internal links and outbound links

Whether a link is a backlink depends on which end of it you stand on. The same <a> element is an outbound link for the site that publishes it and a backlink for the site it points to. Links between pages of one site are internal links.

Internal links are not backlinks, but they are the links you control completely. Google's link best practices say every page you care about should have a link from at least one other page on your site, and that descriptive anchor text on internal links helps both people and Google make sense of the site. The Links report lists them separately, under internal links. If a page matters to you and nothing on your own site links to it, fix that before chasing backlinks; finding the pages that deserve more internal links is usually a quick win.

Outbound links are not a leak either. The same Google page says linking to other sites is nothing to be scared of, and that citing your sources can help establish trustworthiness. It recommends nofollow only when you do not trust the source, not on every external link.

How Google uses backlinks

Backlinks do two jobs in Google Search: discovery and ranking.

Discovery. There is no central registry of web pages, so Google has to keep finding them. Its guide to how Search works calls this URL discovery: some pages are already known, some are listed in sitemaps, and others are found when Google extracts a link from a known page to a new one. Links do most of that work. The SEO Starter Guide says Google primarily finds pages through links from pages it has already crawled, often on other websites, and that the vast majority of the new pages it finds every day are found through links. A new site can still be found through a sitemap, but a link from a page Google already knows is the ordinary way in. The crawl itself is explained in how Google Search works.

Ranking. Google's guide to its ranking systems says it has various systems that look at how pages link to each other, to work out what pages are about and which might be most helpful for a query. PageRank, one of its core ranking systems when Google launched, is among them. Google says how PageRank works has evolved a lot since then and that it is still part of the core ranking systems; as of September 2026 the guide, last updated in December 2025, says the same. Google's page on how it ranks results names one of the factors it uses to judge quality: whether other prominent websites link or refer to the content, which it describes as generally a good sign that the information is trustworthy.

Two things follow. First, links are evidence weighed alongside everything else. The Starter Guide is direct about it: PageRank uses links and is one of Google's fundamental algorithms, but there is much more to Search than links, and PageRank is one ranking signal among many. Second, links do not count equally. The PageRank patent computes a page's rank from the ranks of the pages linking to it, each divided by the number of links on that page, so a page with one link from a very highly ranked page can outrank a page with many links from low-ranked ones. How Google determines a website's authority covers PageRank in more depth, along with the third-party authority scores built on link data.

Because links carry weight, people try to manufacture them: Google's spam policies define link spam as creating links to or from a site primarily to manipulate rankings, and how Google detects and discounts it is covered in quality backlinks in SEO.

What anchor text tells Google

Anchor text, also called link text, is the visible text of a link. Google's link best practices say it tells people and Google something about the page being linked to, and that the words before and after a link matter too, so the whole sentence around a link is part of its meaning. When the link is an image, Google uses the image's alt attribute as the anchor text. When an <a> element is empty, Google can fall back on its title attribute.

Google's guidance on writing anchor text, which applies whether you are linking out or someone is linking to you:

For backlinks, the linking site writes the anchor text, not you. What you do control is what people tend to copy: a clear page title, your organization's name and a page that plainly says what it is. The Links report's top linking text table shows the anchors other sites actually use, which is the quickest way to see how the web describes you. The same commercial phrase repeated across many unrelated sites is a pattern Google's spam policies list as link spam (optimized anchor text in articles, guest posts or press releases distributed on other sites).

The rel attribute: nofollow, sponsored and ugc

The rel attribute on a link describes the relationship between the page and the page it links to. For search, three values matter, all set out in Google's guide to qualifying outbound links:

Four rules go with them, from the same guide, Google's spam policies and its announcements:

• A regular link needs no attribute. Google says so for links you expect it to fetch and parse without qualification.
• Paid links are the one case that is not optional. Google's spam policies accept that buying and selling links is a normal part of advertising, as long as those links carry nofollow or sponsored. sponsored is preferred; nofollow is still acceptable. Since July 2021, Google has also asked sites in affiliate programs to mark affiliate links sponsored, whether they were added by hand or by software.
• Values can be combined, separated by spaces or commas. rel="ugc nofollow" marks a user's link and stays understood by services that only know nofollow.
• Links within your own site call for a different tool. To stop Google fetching one of your own pages, the guide points to a robots.txt disallow rule rather than a rel value.

Here is how the three cases look in markup:

<p>
  We moved our DNS to
  <a href="https://dns.example.net/">a managed provider</a> last year.
</p>
<p>
  This issue is sponsored by
  <a href="https://ads.example.org/offer" rel="sponsored">Example Hosting</a>.
</p>
<p>
  Comment from Sam: I wrote up
  <a href="https://sam.example.com/notes/" rel="ugc nofollow">my own notes</a>.
</p>

Content management systems usually handle the user-generated part, which is what the Starter Guide asks for: every link posted by users should get nofollow or a similar annotation added automatically by the CMS. WordPress core, for example, runs comment text through its wp_rel_ugc function, introduced in version 5.3, which adds rel="nofollow ugc" to every link it finds; the hook that applies it to comments is in WordPress's default filters.

Of the three, only nofollow is part of the HTML standard: the WHATWG HTML Living Standard defines it as a link that the page's author or publisher does not endorse, or that is there mainly because of a commercial relationship. sponsored and ugc are Google's 2019 additions and are not among the link types the standard defines. Two values often seen on links, noopener and noreferrer, are browser instructions rather than search signals: noopener stops a page opened in a new tab or window from getting a reference to the page that opened it, and noreferrer also stops the browser from sending the Referer header, so the site you link to cannot see where the visitor came from. Neither is among the values in Google's guide to qualifying links.

Finally, nofollow works at two levels. On a single link it is a rel value. For a whole page it is a robots rule: <meta name="robots" content="nofollow">, or the same rule in an X-Robots-Tag header, asks Google not to follow any link on the page, and none is shorthand for noindex, nofollow.

How Google treats nofollow, sponsored and ugc

As hints, weighed with other signals. It was not always so.

Google introduced nofollow on January 18, 2005 to fight comment spam: people were posting links in blog comments to raise their own sites' rankings. From then on, links marked rel="nofollow" got no credit when Google ranked sites, and MSN Search and Yahoo supported the initiative. It soon became one of Google's recommended ways to flag paid links too.

On September 10, 2019, Google added sponsored and ugc and changed what all three mean. Instead of excluding marked links outright, Google treats sponsored, ugc and nofollow as hints about which links to consider or exclude, used along with other signals. Its reasoning: links carry useful information, such as how the words in a link describe the page it points to, and looking at every link helps it understand unnatural linking patterns. Google also said that in most cases the change would not alter the outcome: it would generally keep treating ugc and nofollow links as it had treated nofollow before, and not consider them for ranking.

The second step came on March 1, 2020, and it is the one that still catches people out. Some sites relied on nofollow to keep the pages it pointed to out of Google's index. Google said that was never recommended, and since that date nofollow has been a hint for crawling and indexing too. As of September 2026, its guide to qualifying links says marked links will generally not be followed, but that the pages behind them can still be found another way, through a sitemap or a link from another site, and crawled.

What the hint model means in practice:

1. For links you publish, only paid links must be qualified. Google said there is no wrong attribute except in the case of sponsored links: marking a user's link or an ordinary link sponsored costs, at most, that the link may not count as credit for the other page. An ad without sponsored or nofollow is the mistake that risks a link scheme action.
2. For links you receive, a nofollowed or sponsored backlink usually brings no ranking credit, but it is not worthless: people click it, it can send customers, and Google can still read it.
3. Old nofollow links can stay. Google said existing nofollow markup keeps working and there is no need to change it, and suggested moving paid links to sponsored when convenient.

Nofollow vs dofollow links

"Dofollow" is SEO slang for a normal link: one with no nofollow, sponsored or ugc value. There is no dofollow attribute. The HTML standard's list of link types has no such value, and Google's documentation says a regular link needs no rel attribute at all, so there is nothing to add to make a link "dofollow". Asking a site for a dofollow link simply means asking for a normal, unqualified one.

For your own site the rule of thumb follows from the table: leave editorial links alone, mark paid and affiliate links sponsored, let your CMS mark user links ugc, and keep nofollow for the rare link you need to publish but do not trust.

How to see your backlinks in Search Console

Google's free view of your backlinks is the Links report in Search Console. Once your site is verified as a property:

1. Open Links in the Search Console menu.
2. Under External links, read the three tables: Top linked pages (your pages with the most backlinks), Top linking sites (the root domains that link to you most) and Top linking text (the anchor text other sites use).
3. Select More under a table for the full list, then select a page or a site to drill down to the individual linking pages.
4. Use the Export button on the report's main page to download Latest links, the most recently discovered links sorted by date, or More sample links, a sample drawn from all the links Google knows. Each export holds up to 100,000 rows.
5. Check the internal links table as well: it shows which of your pages your own site links to most.

Google's suggested uses are practical: confirm that your most linked pages are the ones you expect, that you recognize the sites linking to you, and that the link text describes your site accurately.

Know what the report is not:

• It is a sample. Google says it is not a comprehensive list of every link, and some URLs are left out, for example non-indexed pages.
• Tables stop at 1,000 rows, so large or heavily linked sites see truncated tables.
• Sites are grouped by root domain, with protocol, subdomains and subfolders folded together.
• Your pages are grouped by canonical URL. A link to an old address that redirects, or to a duplicate Google has canonicalized, is shown against the target page.
• It includes links found over time, some of which may since have been removed.
• It does not say whether a link is nofollow, and it shows no score of any kind.

Third-party tools see a different sample. Moz's Link Explorer and Ahrefs, for example, report backlinks from their own indexes: Moz notes that updates to its web index can include more or fewer linking domains than earlier crawls, and Ahrefs scores a site against the others in its own database. Expect their counts to differ from Search Console's and from each other's; their authority scores are explained in how Google determines a website's authority. For links from sites that look like spam, the Links report help says Google typically ignores links from known spam sites; when to act on them, and when the disavow tool is appropriate, is covered in quality backlinks in SEO.

Backlink myths, checked against Google's documentation

"Power backlinks" and the other labels for strong links are discussed in quality backlinks in SEO. If you would rather have someone check how your site's links are crawled and reported, our SEO service starts with a technical audit of crawling, rendering and indexing, fixes internal links as part of on-page work and reports from Search Console. It earns links with content worth citing and digital PR where it fits, and does not buy links or use networks that break search engine guidelines.