什么是 Nofollow 链接？定义与指南

Nofollow 链接是指示搜索引擎不要传递排名权重的超链接。

SEO and Search

A no-follow link is a hyperlink that carries rel="nofollow", a note telling search engines the publisher doesn't vouch for the page on the other end. The link still works for people, because the value labels a relationship rather than locking a door.

That distinction has mattered more since 2019, when Google stopped treating nofollow as a strict command and started reading it as a hint. Most confusion comes from expecting it to block crawling, indexing, or ranking. It doesn't reliably do any of those. If you're looking at this from the receiving side, our guide to how backlinks work covers the basics.

What rel="nofollow" Actually Signals

The HTML meaning: a non-endorsement note

Start with the standard, not with search engines. The HTML Living Standard's link types table defines nofollow as a signal that the document's original author or publisher "does not endorse the referenced document". That's the whole definition. It says nothing about crawlers, rankings, or indexes.

Nofollow describes a stance toward a destination. Clicking still works, the browser still navigates, and the reader sees an ordinary link. Everything a search engine does with the value is its own interpretation, layered on top of that original meaning.

Popular references don't always keep up. The lead sentence of Wikipedia's Nofollow reference article still describes a setting that "directs search engines not to use the link for page ranking calculations", which reads closer to the original 2005 rule than to Google's current wording; the article covers the hint change further down.

Why "dofollow" isn't a real attribute

You'll hear people talk about "dofollow links" as if they were a separate type. They aren't. Google's guide on qualifying outbound links says regular links don't need a rel attribute at all. "Dofollow" is just shorthand for a link with no qualifying value, and writing rel="dofollow" adds nothing. The backlink entry linked above covers the receiving-site view.

How Google Reads Nofollow Today

From comment-spam fix to hint

Nofollow began as a spam control. Google announced it in Google's January 2005 comment spam post, titled "Preventing comment spam". As the title says, the original target was spam left in comment sections.

The value also has an open-spec history. The original rel-nofollow microformat specification, dated 2005-01-10 and edited by Tantek Çelik and Kevin Marks of Technorati, said a hyperlink's destination "SHOULD NOT be afforded any additional weight or ranking". Put that next to the HTML standard's non-endorsement wording and you can see the meaning drift. It started as a ranking instruction. Today it reads as a relationship label.

The bigger shift came in September 2019. In Google's 2019 nofollow announcement, Danny Sullivan and Gary Illyes introduced rel="sponsored" and rel="ugc" and explained that nofollow, sponsored, and ugc would all be treated as hints about which links to consider or exclude. For ranking, the hint model applied from the announcement onward. For crawling and indexing, it took effect on March 1, 2020.

Trade press read the change the same way at the time. Search Engine Journal's February 2020 explainer noted that Google previously treated nofollow links as a directive, and it pointed site owners toward meta robots noindex as the better way to keep pages out of the index.

The September 2019 post on the Google Search Central Blog also carries a line worth remembering: "...no wrong attribute except in the case of sponsored links."

In other words, Google doesn't treat a nofollow and ugc mix-up as an error. Paid placements are the exception it singles out.

What "generally not followed" means

Google's current wording is careful. Its guide to qualifying outbound links in Google Search says links marked with these rel values "will generally not be followed". Generally isn't always. The same page reminds site owners that a linked page can still be discovered elsewhere, for example through a sitemap or another site's link.

So a nofollow link doesn't hide its destination. If any other path points to that URL, a crawler can still reach it. Treat nofollow as a statement about your link, not as a fence around somebody else's page.

PageRank sculpting and why it stopped

Some sites once added nofollow to internal links to push ranking value toward priority pages, a practice called PageRank sculpting. In Matt Cutts's 2009 sculpting post, Cutts, then at Google, wrote that nofollow links didn't pass PageRank or anchor text, explained that the share assigned to nofollowed links wasn't handed on to the remaining links, and said he wouldn't recommend sculpting.

That's 2009 behavior, since replaced by the hint model. The lesson still holds. Nofollow on internal links isn't a lever for steering ranking value. If the goal is crawl control on your own site, Google points to a robots.txt disallow rule, and if the goal is keeping a page out of results, it points to noindex.

Nofollow, Sponsored, or UGC: Picking by Relationship

Google's values map to three relationships. Pick the one that describes how the link got there, then stop.

Paid placements and affiliate links

Mark advertisements and paid placements with rel="sponsored". The Qualify page says nofollow is still an acceptable way to flag paid links, though sponsored is preferred. Google's spam policies on links, last updated on 28 Aug 2026, go further, describing buying and selling links as a normal part of the web's economy and stating that such links don't violate policy as long as they're qualified with nofollow or sponsored.

Affiliate links get one relevant mention. In its site reputation abuse section, the same policy page lists affiliate links used throughout a page, "with links treated appropriately," as consistent with that policy. We won't stretch that into rules the documentation doesn't state. One analysis point, not policy: a rel value speaks to machines, so it doesn't replace telling human readers that a link is paid.

User-submitted links

Comments and forum posts are the textbook cases of user-generated content. Google recommends rel="ugc" for these, and Google's documentation on crawlable links says that if users can insert links on your site, you should add ugc or nofollow to those links too. Applying it at the template level is the practical route.

Editorial links to sources

Many sites overcorrect at this point. When you choose a link yourself because the destination helps your reader, it doesn't need a qualifying value. Google's crawlable links page advises reserving nofollow for destinations you don't trust instead of automatically applying it to all external links. The same page notes that external links can help establish trustworthiness, for example when you cite your sources.

Write clear, descriptive anchor text for each link, leave the rel attribute off, and let the link do its job. Blanket nofollow is a habit, not a policy.

Nofollow vs. Noindex vs. Robots.txt

These three get mixed up because each sounds like it controls search visibility. They work at different levels. Nofollow qualifies a single link. Noindex and robots.txt rules govern your own URLs.

Keeping a page out of the index

Nofollow on links pointing at a page won't keep it out of search results, since it can be discovered elsewhere. Google's instruction is specific: let the page be crawled, then apply the noindex robots rule. That's why crawling has to stay open. The rule only works if the page can be read.

Controlling crawl on your own site

If you need to stop Google from fetching a page on your own site, nofollow on your internal links is the wrong tool. Google says to use the robots.txt disallow rule instead. Keep each tool on its own job:

Nofollow, sponsored, ugc: describe your relationship to a link's destination.
Noindex: keeps a page you control out of the index, provided it can be crawled.
Robots.txt disallow: tells crawlers not to fetch specific URLs on your site.

Implementing and Auditing Nofollow Markup

Markup Google can read

The attribute only helps if Google can read the link. Google's crawlable links documentation says it can't reliably extract URLs from <a> elements that don't have an href attribute. So the basic pattern is a real anchor with a real destination:

<a href="https://example.com/offer" rel="sponsored">Partner offer</a>

You can also combine values. The Qualify page says multiple rel values may be given as a space- or comma-separated list, so a paid link that a user submitted could carry rel="ugc sponsored". After publishing, inspect the rendered HTML rather than trusting a CMS checkbox. Templates, plugins, and scripts can all change what actually ships.

A relationship audit for existing links

Most sites inherit years of link markup that nobody remembers deciding on. A short audit fixes that. For each outbound link, ask three questions:

Who chose this link? An editor, a user, or a paying partner.
Was anything exchanged for it? Payment of any kind counts here.
Would you vouch for the destination? If not, and nothing else applies, that's nofollow territory.

The answers map straight onto the values. Paid placements get sponsored. User-submitted links get ugc. Links you'd rather not vouch for, where neither of those applies, get nofollow, and editorial links you'd stand behind get nothing at all. If your team buys placements alongside search engine marketing programs, those paid links are the first batch worth checking. Log the source page, destination, who chose it, any exchange, rendered rel values, and check date, so the next review starts from evidence.

Judging an Incoming Nofollow Backlink

Referral value you can measure

A nofollow link from a relevant site is still a working link. Readers can click it, land on your page, and convert. That value shows up in your own analytics as referral sessions, engaged visits, inquiries, and sales, and none of it depends on how a search engine weighs the link.

Judge the placement the way you'd judge any channel. Does it reach the right audience? Does it send qualified visits? Do those visits lead somewhere useful for the business? A nofollow placement sending steady, relevant traffic can beat a followed link nobody clicks, but only your own numbers will tell you.

What you can't infer about ranking

Be honest about the limit here. Because Google treats nofollow as a hint, there's no public formula for how much ranking credit, if any, a particular nofollow link receives. The Qualify page says such links will "generally" not be followed. That word leaves room for discretion, but it doesn't promise a benefit. Don't report a nofollow link as a ranking win or write it off as worthless; report what you observed, and measure rankings within your wider search engine optimization program.

No-Follow Link FAQs

Does a nofollow link pass any ranking value?

Google doesn't publish a figure. Since 2019 it has treated nofollow as a hint rather than a command, and its current documentation says these links will generally not be followed. Plan for referral value, not ranking credit.

Is rel="sponsored" required for paid links?

It's preferred, not required. Google accepts nofollow on paid links too, but sponsored is the value it recommends, and paid links are the one case where Google says the attribute choice can be wrong.

Can I combine rel values on one link?

Yes. Google accepts multiple values in a space- or comma-separated list, such as rel="ugc sponsored" for a paid link that a user submitted.

Does nofollow work on internal links?

It isn't the right tool. Using nofollow on internal links to steer ranking value is called PageRank sculpting, and Matt Cutts advised against it in 2009. For crawl control, use a robots.txt disallow rule, and for index control, use noindex.

Is "dofollow" a real attribute?

No. Regular links don't need any rel value, according to Google's Qualify page. "Dofollow" is informal shorthand for a link without a qualifying value.

Does nofollow matter for AI search?

We haven't found documentation from AI providers on how they treat rel="nofollow", so we won't guess. One related point is documented: Google's current spam policies page now counts attempts to manipulate generative AI responses in Google Search as spam.

Related Terms

Link Building in SEO - earning links from other sites, where qualification decisions start.
Black Hat SEO tactics - manipulative practices, including unqualified paid links.
Dofollow Link - informal name for a link with no qualifying rel value.
PageRank - Google's original link-based ranking signal.
Noindex - the robots rule that keeps a crawlable page out of the index.