Nofollow is a value of a link's rel attribute that asks Google not to count that link as a vote for the destination page.
What to know before using it: since 1 March 2020 nofollow has been a hint, not an order. Google may decide to follow the link anyway, and the page you linked to can end up in the index without the attribute stopping it. If you need to keep a page out of the results, nofollow is not the right tool.
How to write it
<a href="https://example.com" rel="nofollow">link text</a>It is not a tag of its own: it is a value of the rel attribute, the same one that also holds sponsored, ugc and noopener. Several values can sit together separated by a space, so rel="nofollow ugc" is correct, and Google suggests it to anyone who wants to stay compatible with services that do not read the attributes introduced in 2019.
The opposite, "dofollow", does not exist. A link with no rel attribute is already followable and needs nothing. Google ignores rel="follow" and rel="dofollow" because they are not official attributes: the HTTP Archive Web Almanac still finds them on more homepages than the two real attributes.
Which value to use, according to Google
Google Search Central's documentation on outbound links assigns a specific value to each situation.
| Value | When Google says to use it | Typical example |
|---|---|---|
rel="sponsored" | Links that come from advertising, sponsorships or other compensation agreements | Sponsored article, banner, affiliate link |
rel="ugc" | Links inside user-generated content | Comments, forum posts, reviews |
rel="nofollow" | When you do not want to imply any endorsement, including passing ranking credit | A source you cite but do not vouch for |
no rel attribute | Standard links you want crawled and analysed with no qualifying criteria | Citing a source you endorse |
For paid links Google prefers sponsored, but says nofollow is still an acceptable way to flag them. What you cannot do is leave them unmarked: a bought link left followable falls under the link schemes the spam policies penalise, and it is one of the quickest ways to slide into black hat SEO without meaning to, while believing you are doing white hat SEO.
From directive to hint: what changed in 2019
On 10 September 2019 Danny Sullivan and Gary Illyes announced on the Google Search Central blog the two new attributes and a change to nofollow, which had existed since January 2005, when it was introduced to fight comment spam. The wording is clear: all three link attributes, sponsored, UGC and nofollow, are "treated as hints about which links to consider or exclude within Search".
The stated reason is that links carry useful information regardless: for example the way the anchor text describes the page it points to, and the fact that looking at all links helps spot unnatural linking patterns. For ranking the change took effect immediately; for crawling and indexing nofollow became a hint on 1 March 2020.
In the same post Google warned anyone relying on nofollow alone to stop a page being indexed, something that was "never recommended", to switch to more robust mechanisms. There are two, and they do different things: noindex removes the page from the index, a block in robots.txt stops the crawler from downloading it.
On a single link or on the whole page
Nofollow also exists as a robots meta tag rule, and in that form it does not apply to one link but to every link on the page, internal ones included.
<meta name="robots" content="nofollow">The official robots meta tag specification describes it like this: "Do not follow the links on this page. If you don't specify this rule, Google may use the links on the page to discover those linked pages." The noindex, nofollow combination has a shortcut, none, and when two rules conflict the more restrictive one wins.
In real projects the nofollow meta robots is rarely useful. It also switches off the internal paths that lead the crawler to your good pages, and nearly always whoever set it wanted to deal with three external links, not all of them.
How much it is really used
The 2022 Web Almanac, which analyses more than eight million HTTP Archive homepages worldwide, found rel="nofollow" on 29.5% of mobile homepages, down 1.2 points on the year before. The attributes introduced in 2019 have barely moved: rel="ugc" appears on 0.4% of pages, rel="sponsored" on 0.5% of desktop and 0.4% of mobile pages.
On the backlink side the reference is still Ahrefs' 2020 study of the 110,000 strongest domains in its index, also global: 10.6% of backlinks to those sites were nofollow, while ugc showed up on 0.44% of referring domains and sponsored on 0.01%. These are 2020 figures, but the Web Almanac survey two years later tells the same story: the attributes introduced in 2019 never took off.
When the link you receive is nofollow
WordPress marks links in comments automatically, and so does MediaWiki, the software behind Wikipedia, which adopted nofollow right after the 2005 announcement. This means a sizeable share of the mentions a site picks up naturally arrive with the attribute already on, and a profile made only of followable links looks very little like one that grew on its own.
In link building this changes how you judge an opportunity. A nofollow link from a publication your customers read brings visits and puts your name in a relevant context; a followable link from a directory nobody reads brings nothing measurable. In the projects we handle as SEO consultants, the first question about a link is never whether it is follow, but where the traffic of the page hosting it comes from.
The habit that costs more than doing nothing
That leaves nofollow on internal links, added to "avoid wasting authority" on utility pages. Matt Cutts, then head of Google's webspam team, took the mechanism apart in June 2009 with a simple sum: a page with ten points of PageRank and ten outgoing links, five of them nofollow. Before, the five normal links got two points each; after the change they get one. The five points meant for the nofollow links are not redistributed to the others, they are simply lost.
Eleven years later Ahrefs still found internal nofollow on 35.3% of the sites it analysed. It is the rare case of an optimisation that costs more than doing nothing: whoever leaves things alone gets the better result.
Nofollow FAQs
No. Since 1 March 2020 nofollow has also been a hint for crawling and indexing: Google can follow the link anyway and the destination page can appear in the results. To keep it out of the index you need noindex; to stop it being crawled, a block in robots.txt.
No. HTML has no rel="dofollow" value: a link with no rel attribute is already followable and needs nothing. Google ignores both rel="follow" and rel="dofollow" because they are not official attributes.
Google prefers rel="sponsored", but its documentation still treats nofollow as acceptable for flagging paid links, with the same effect for this purpose. You can also combine them. The mistake to avoid is leaving the link with no attribute at all.
As a rule, no. The PageRank meant for a nofollow internal link is not redistributed to the other links on the page: it disappears. Matt Cutts explained this in June 2009 and Google has not announced any later change to the calculation.
It brings the visits of people who click on it and a mention of your name in the context where it appears. On rankings, Google says it treats the attribute as a hint and generally does not count these links for ranking purposes.
