How Google evaluates links today

What Google officially says about links

Google's link documentation opens with one sentence that sums up the whole topic: “Google uses links as a signal when determining the relevancy of pages and to find new pages to crawl.” So a link does two jobs. It is a path that Googlebot follows to discover a URL, and it is information about the page it points to.

The ranking systems guide describes the second job in more detail:

Note the wording: “various systems”, and the purpose is to understand what pages are about and which are helpful. Google does not describe links as votes to be counted. In a 2021 blog post it put the same idea from the site owner's side:

“One of the ways” is important. Links are a signal among many, not the whole ranking. A page with strong links and weak content does not win by default, and a page with few links can rank well when it is the most helpful answer.

PageRank today

PageRank is the link analysis method Google was built on. The ranking systems guide says only this about its current role, and it is worth reading exactly:

That is the documented position: PageRank still exists, it is part of the core systems, and it has changed a lot since the original paper. Google does not publish how it is calculated today, how much it weighs, or a public PageRank score for any page. Any tool that shows you a “PageRank” number is showing its own estimate.

The concept also appears in Google's practical help. The Manual actions report tells site owners to ask for spammy links to be removed or changed so they no longer “pass PageRank”, for example with rel="nofollow" or a more specific value. In other words, Google still thinks of links as something that can pass ranking credit, and of rel values as a way to stop that.

How a link becomes a signal

Before a link can mean anything, Google has to find it, read it and decide how much to trust it. The diagram below is a simplified model of that path. Google does not publish its pipeline, so treat the stages as an illustration, not documentation.

Each stage is grounded in something Google documents:

• Crawlable. “Generally, Google can only crawl your link if it's an <a> HTML element (also known as anchor element) with an href attribute.” A link built with a <span> or only a JavaScript click handler may never be extracted.
• Understood. About anchor text, Google says: “This text tells people and Google something about the page you're linking to.” The surrounding words matter too, which is why Google advises against chains of links with no context.
• Qualified. The rel value tells Google about your relationship with the linked page. Since 2019 these values are hints, covered below.
• Trusted. Spam detection systems try to identify links that exist to manipulate rankings and neutralise them.

The technical side of anchors and rel markup is covered in depth in the lesson Links on the page: anchors and rel values. This lesson is about how those links are judged.

What link spam is

Google's spam policies define it in one sentence:

Two words do the work. “To or from” means selling links is covered as much as buying them. “Primarily” means the test is the purpose of the link: a link that exists mainly to move rankings is spam, whatever it looks like. Google then lists examples. The cheat sheet below gives each one in Google's exact words, with a typical case from practice.

Paid links are not banned as such. Google is explicit:

So the line is not money but disclosure. An advertisement with rel="sponsored" is fine; the same advertisement as a plain followed link is link spam. The same applies to affiliate links: “We ask sites participating in affiliate programs to qualify these links with rel="sponsored", regardless of whether these links were created manually or dynamically.”

How Google neutralises spammy links

Google fights link spam in two ways. The spam policies say: “We detect policy-violating practices both through automated systems and, as needed, human review that can result in a manual action.” The automated part is where most of the work happens.

SpamBrain and link spam updates

In July 2021 Google launched a “link spam update” and wrote that it had “gotten significantly better at detecting and nullifying such link schemes”. In December 2022 it applied SpamBrain to links:

The same post explains what happens to the links it catches:

What “neutralised” means for you

The key word is neutralised. Google's stated goal is that manipulative links simply stop counting: “Our algorithms and manual actions aim to nullify these unnatural links at scale”. In practice this has three consequences:

• Bought links often stop working without warning. If rankings were propped up by paid links, they can fall when those links are discounted, even with no manual action and no message in Search Console.
• Random spam pointing at your site is usually ignored. Most sites collect scraper and directory links they never asked for. In most cases you do not need to do anything about them.
• Selling links costs you too. SpamBrain also looks for sites “used for the purpose of passing outgoing links”, so a site that sells followed links can lose the trust of its own outgoing links.

Google's broader policy statement still applies: “Sites that violate our policies may rank lower in results or not appear in results at all.” Neutralising is the common outcome, not the only possible one.

nofollow, sponsored and ugc are hints

Most links need no attribute at all: “For regular links that you expect Google to fetch and parse without any qualifications, you don't need to add a rel attribute.” For the rest, Google offers three values:

Until 2019, Google did not count nofollow links as a signal at all. That changed:

Google's current documentation adds that “Links marked with these rel attributes will generally not be followed.” Read the two statements together: a qualified link is normally excluded from ranking credit, but Google may still look at it, for example to understand linking patterns. Do not expect nofollow links to work like followed ones, and do not assume they are invisible either. Values can be combined, such as rel="ugc nofollow".

<!-- paid placement or affiliate link -->
<a href="https://example.com/product" rel="sponsored">Example product</a>
<!-- link in a reader comment -->
<a href="https://example.com/forum" rel="ugc nofollow">forum thread</a>

Which links actually help

Google does not publish a list of link qualities or their weights. What follows is practice: patterns that SEO teams consistently see and that agree with how Google describes links. Treat it as judgement, not as a formula.

• Relevance. Google uses links to understand “what pages are about”. A link from a page about running gear to your running-shoe guide says something about your topic. A link from an unrelated casino page says very little.
• Editorial placement. A link someone chose to put in the main text, because it helps their reader, is the opposite of the spam list. Footer, sidebar and template links repeated across many sites look like the “widely distributed” pattern Google names.
• A real audience. A page that people actually read, on a site with its own readership, is the kind of place a well-deserved link comes from. If a page exists only to host links, it fits Google's “low-value content” example.
• Natural anchor text. Brand names, page titles and descriptive phrases are normal. A profile full of exact-match commercial anchors is what “optimized anchor text” in Google's policy describes.
• Crawlable markup. A plain <a href> that Google can parse. A link Google cannot extract cannot count.

As a rule of thumb, one relevant editorial link from a real publication in your field usually does more than many directory, profile or comment links. Google's own summary points the same way: “Focusing on producing high quality content and improving user experience always wins out compared to manipulating links.” Ways to earn such links, from original data to expert commentary, are covered in the lesson Digital PR and brand mentions.

What to ignore: DA, DR, TF and other third-party scores

Domain Authority (Moz), Domain Rating (Ahrefs), Trust Flow (Majestic) and Authority Score (Semrush) are third-party estimates. Each vendor calculates its score from its own crawl of the web and its own formula. Google's documentation does not mention any of them, and Google does not use them. Raising your DA does not raise your rankings; at most, both can move together because both react to the same real links.

These scores are still useful, as long as you use them for what they are:

• to sort a long list of prospects quickly before reviewing them by hand;
• to compare sites roughly, using one vendor's score consistently;
• to spot sudden changes in your own profile that deserve a closer look.

What they should not be is a target or a price list. A site can have a high score because it sells links to everyone, which is exactly what SpamBrain is designed to detect. A small, focused site in your niche with a modest score can be the better link. Judge the page, its topic and its readers, not the number.

Myths about links

Internal vs external links

Everything Google says about crawlable markup and anchor text applies to links inside your site as well. The difference is control: you cannot decide who links to you, but you decide every internal link.

For internal links Google's advice is direct: “Every page you care about should have a link from at least one other page on your site.” It also refuses to give a quota: “There's no magical ideal number of links a given page should contain.” Use descriptive anchors and link where it helps the reader. Internal links cannot be link spam in the sense of the policy as long as they are not hidden or stuffed with keywords, but they cannot replace the outside signal that other sites choose to link to you either.

For links you place to other sites, Google takes away a common fear:

And it warns against the opposite habit: “Use nofollow only when you don't trust the source, and not for every external link on your site.” Adding nofollow to every outbound link to hoard PageRank is not a Google recommendation. Structure, depth and orphan pages are covered in the internal linking lessons of this course.

The risk side: manual actions for unnatural links

When automated neutralising is not enough, a human reviewer can apply a manual action. You will see it in the Manual actions report in Search Console. Two types concern links.

Unnatural links to your site

Google's description: “Google has detected a pattern of unnatural, artificial, deceptive, or manipulative links pointing to your site.” It warns: “This can result in some or all of your site getting a manual action.” The recommended steps are to export your links from Search Console, find the ones that break the link spam policy, ask site owners to remove them or qualify them, disavow what you cannot remove, and then request a review with documentation.

Google is strict about the disavow step: “Blindly adding all backlinks to the disavow file is not considered a good-faith effort, and will not be enough to make your reconsideration request successful.”

Unnatural links from your site

This action targets the seller side: a pattern of paid or exchanged outbound links on your site. The fix is to remove those links or make them stop passing PageRank, for example with rel="sponsored" or rel="nofollow", and then request a review.

How to export links, decide what is really unnatural and write a disavow file is covered step by step in the lesson Link audits and disavow.

Common mistakes

• ✕Buying followed links and calling them “guest posts”If payment or a product was involved, the link needs rel="sponsored".
• ✕Exact-match anchors everywhereOptimised anchors in guest posts and press releases are a named spam pattern.
• ✕Choosing link sources by DA or DR aloneCheck topic, placement and real readers instead.
• ✕Panicking about random spam linksGoogle aims to nullify them; disavowing organic links can hurt you.
• ✕Selling links on your own site without qualifying themThis can trigger “Unnatural links from your site”.
• ✕Footer and widget credit links with keywordsWidely distributed template links are on Google's list.
• ✕nofollow on every outbound linkGoogle advises using it only when you do not trust the source.
• ✕Links Google cannot parseButtons and JavaScript handlers without href pass nothing.

FAQ

Sources

• Google Search Central — Link best practices for Google
• Google Search Central — A guide to Google Search ranking systems
• Google Search Central — Spam policies for Google web search
• Google Search Central — Qualify your outbound links to Google
• Google Search Central Blog — Evolving “nofollow” (2019)
• Google Search Central Blog — A reminder on qualifying links and our link spam update (2021)
• Google Search Central Blog — December 2022 link spam update
• Google Search Console Help — Manual actions report

Quotes checked against the official pages on 30 October 2026.