Nofollow is a value placed in a hyperlink's rel attribute, written rel="nofollow", which tells search engines that the publishing site does not vouch for the destination and that the link should not count towards that destination's ranking. It exists because Google's results were built on PageRank, a score calculated from links between pages, and spammers were flooding blog comments, guestbooks and referrer lists with links to inflate it. Google introduced the value on January 18, 2005 to make those links worthless. Two decades on, it is also the standard way to disclose paid and affiliate links to search engines, and since September 10, 2019 Google has read it as a hint rather than a command.
How the value works
The attribute sits inside the anchor element of HyperText Markup Language (HTML), the tag that creates a link. An unqualified link passes ranking credit by default. Since 2019 Google has recognised three qualifying values: sponsoredfor advertisements and paid placements, ugc for user-generated content (UGC) such as comments and forum posts, and nofollow for any other link a site prefers not to be associated with. Values can be stacked, so rel="ugc nofollow" is valid. There is no "dofollow" value; a link without a qualifier is simply an ordinary link.
The same word also works across a whole page. Placed in a robots meta tag, or sent in an X-Robots-Tag Hypertext Transfer Protocol (HTTP) header, nofollow covers every link on the page. That vocabulary predates the link attribute by almost nine years: the robots meta tag came out of a World Wide Web Consortium (W3C) workshop in May 1996, with follow and nofollow among its six terms. Google's documentation shows an Apache server rule that attaches "noindex, nofollow" to every Portable Document Format (PDF) file on a site, and Apple's crawler, Applebot, disregards all links on a page tagged nofollow.
Each engine decides what the label means. As of October 2026, Google's documentation says links carrying any of the three values are generally not followed, but the destination can still be found through sitemaps or other sites' links; robots.txt blocks crawling and noindex keeps pages out of results. When Google expanded its guidance on filter pages in online shops in December 2024, it noted that nofollow on faceted navigation links can influence crawling, while presenting robots.txt rules and web address (URL) fragments as firmer tools.
The arithmetic is unforgiving. Matt Cutts, then head of Google's webspam team, explained on June 15, 2009 that a page with five ordinary and five nofollowed links divides its PageRank ten ways, and the five nofollowed shares are lost rather than redistributed. The change, made more than a year earlier, undermined PageRank sculpting, the practice of nofollowing internal links to concentrate credit on chosen pages.
Responsibility splits along the supply chain. On the sell side, publishers set the value through content management systems. WordPress has nofollowed commenters' links since version 1.5 and, since version 5.3 shipped on November 12, 2019, writes rel="nofollow ugc" on them by default. English-language Wikipedia has nofollowed external links from its articles since January 20, 2007. On the buy side, advertisers and agencies commissioning sponsored articles, affiliate deals and digital public relations (PR) check placements for a qualifier.
Programmatic display rarely needs one. In a 2011 video, Cutts said most banner ads sold through exchanges and ad networks click through redirects that crawlers are already blocked from. The exposure sits in native formats: Google's spam policies class paid advertorials whose links pass ranking credit as link spam, alongside text ads that fail to block it.
From comment spam to hints
Cutts announced the value with Jason Shellen, Blogger's programme manager, writing that marked links "won't get any credit" in ranking. Blog software makers including WordPress and Six Apart signed on that day, and MSN Search and Yahoo backed it. The HTML Living Standard now defines the keyword as a statement that the author or publisher does not endorse the linked document, according to Mozilla's MDN Web Docs.
Its purpose widened quickly. In December 2007, Cutts and Maile Ohye wrote that paid links should be disclosed through nofollow or through a redirect via a page blocked in robots.txt. On March 11, 2016, Google told bloggers receiving free products to nofollow links to suppliers. On July 26, 2021, it asked affiliate sites to use rel="sponsored" and began a link spam update that rolled out over two weeks.
The pivot came on September 10, 2019. Google added sponsored and ugc and said all three values would be treated as hints for ranking; for crawling and indexing, nofollow became a hint on March 1, 2020. Links hold useful information, such as the words describing their destination, and seeing every link helps expose unnatural patterns, Google explained. Bing's Fabrice Canel said Bing had always read nofollow as a hint; its guidelines added sponsored and ugc on June 30, 2020, though a spokesperson told Search Engine Land the newer values were a weaker signal.
Enforcement hardened alongside. In December 2022 Google deployed SpamBrain, its machine learning spam system, against sites buying links and sites selling them. Since then Google has asked that paid and affiliate links carry sponsored or nofollow, and describes any ranking benefit from neutralised links as permanently lost.
Why it matters for marketers
Links still carry commercial weight. A survey of 131 search practitioners released on September 9, 2026 placed backlinks second among 103 factors, at 54.8%. Google's own use of them is narrower than that suggests: a federal court memorandum filed on September 2, 2025 in the US antitrust case described PageRank as one component of Google's quality score, with most quality assessment drawn from page content.
For anyone buying or selling placements, the qualifier separates an advertising transaction from a policy breach. Google's spam policies accept paid links as a normal part of the web's economy, provided they carry nofollow or sponsored. Unqualified, they can trigger a manual action for unnatural links pointing in or out, or be neutralised algorithmically. When Google expanded its site reputation abuse guidance in early 2025, publishers moving third-party sections to other domains were told that links between the sites must carry nofollow.
The qualifier is not disclosure to readers. The US Federal Trade Commission's (FTC) revised endorsement guides, effective July 26, 2023, treat an affiliate link as a material connection that must be disclosed clearly and conspicuously; a rel value is invisible to people. Nor does using it carry a penalty: Google's Martin Splitt said in August 2024 that nofollow does not signal low quality.
Limits and disputes
Nofollow did not end comment spam. On January 18, 2026, more than 21 years after launch, Google's John Mueller was still telling a worried site owner that links dropped into comments by spammers had no effect.
The design drew objections from the start. In January 2005 the web developer Lachlan Hunt called the name "an extremely poor choice" on a W3C mailing list, arguing that it described the behaviour of one kind of software rather than a relationship between documents.
Blanket use creates collateral damage. When Wikipedia nofollowed its external links at Jimmy Wales's request, the blogger Philipp Lenssen argued that the site now "takes from the communities but doesn't give back", since its own rankings rest partly on the sources it cites.
The hint model adds opacity: no outside party can tell whether Google has counted a particular nofollowed link. The values meant to refine the signal have barely been adopted: the HTTP Archive's 2025 Web Almanac found nofollow on 32.3% of pages, against 0.5% each for sponsored and ugc. Control also runs one way. Pedro Dias, a former Google search quality engineer, said in April 2026 that Google can silently cut a site's ability to pass PageRank, in effect nofollowing it from outside without notice.
Not the same as
Robots meta nofollow covers every link on a page; the rel value works link by link.
Noindex and robots.txt disallow control indexing and crawling. Nofollow does neither reliably.
Noopener and noreferrer are browser values: the first stops a new tab controlling the page that opened it, the second withholds the referring address, which affects analytics attribution. Neither has a search meaning.
Disclosure labels such as "ad" in visible text address readers and regulators; the rel value addresses crawlers. Neither substitutes for the other.
Recent developments
Google's search engine optimisation (SEO) Starter Guide, read alongside a new guide to generative artificial intelligence (AI) search in May 2026, says links posted by users should automatically carry nofollow or a similar annotation. On May 15, 2026, Google extended its spam policies, link spam included, to AI Overviews and AI Mode. The consultant Lily Ray asked whether the accompanying language on inauthentic mentions covered paid brand mentions, a form of promotion with no link and therefore no rel attribute to qualify.
Google revised its site reputation policy on August 28, 2026, and from August 30 its manual actions stopped applying to searches in the European Economic Area (EEA); the revised page still accepts affiliate links handled under the link spam rules. The September 2026 spam update began on September 24 with a rollout window of up to two weeks, and Google has not described it as a link spam update.
At Search Central Live in Barcelona on October 2, 2026, Gary Illyes showed internal timing data in which link annotations typically take from minutes to three weeks, and months at worst, according to a recap by John Campbell of the agency ROAST. Illyes also warned that scaled content was becoming a bigger problem than link spam.
Timeline
- May 1996: A W3C workshop defines the robots meta tag, including follow and nofollow
- January 18, 2005: Google introduces rel="nofollow", backed by MSN Search and Yahoo
- January 20, 2007: English Wikipedia applies nofollow to external links in articles
- December 2007: Google says paid links should be disclosed through nofollow or a robots.txt-blocked redirect
- June 15, 2009: Matt Cutts confirms that PageRank assigned to nofollowed links is lost
- March 11, 2016: Google asks bloggers reviewing free products to nofollow supplier links
- September 10, 2019: Google adds sponsored and ugc and treats all three values as ranking hints
- November 12, 2019: WordPress 5.3 applies rel="nofollow ugc" to comment links by default
- March 1, 2020: Nofollow becomes a hint for crawling and indexing at Google
- June 30, 2020: Bing's revised guidelines add sponsored and ugc
- July 26, 2021: Google asks affiliates to use sponsored and starts a link spam update
- December 14, 2022: Google's link spam update deploys SpamBrain against link buyers and sellers
- July 26, 2023: The FTC's revised endorsement guides take effect
- August 2024: Martin Splitt says nofollow does not signal low quality
- December 2024: Google publishes expanded faceted navigation guidance
- September 2, 2025: A US court memorandum describes PageRank as one component of Google's quality score
- January 18, 2026: John Mueller says spam comment links have no ranking effect
- May 15, 2026: Google applies its spam policies, link spam included, to AI Overviews and AI Mode
- August 28, 2026: Google revises its site reputation policy, with EEA changes effective August 30
- September 9, 2026: A survey of 131 practitioners ranks backlinks second among 103 factors
- September 24, 2026: Google begins the September 2026 spam update
- October 2, 2026: Gary Illyes presents link annotation timing data in Barcelona
Related PPC Land coverage
- Explaining nosnippet - The robots meta tag's 1996 origins and the vocabulary it shares with nofollow.
- Google adds AI Mode to robots meta tag documentation - Page-level directives, including server rules applying noindex and nofollow to PDFs.
- Apple unveils details about Applebot - How Apple's crawler reads nofollow and other robots meta values.
- Managing faceted navigation URLs: new Google documentation - Google's December 2024 guidance on filter links, crawling and nofollow.
- Google activates AI SpamBrain against link spam - The December 2022 link spam update targeting link buyers and sellers.
- Explaining SpamBrain - Google's spam system, its treatment of paid links and the permanence of neutralised link value.
- 131 SEOs rank relevance top Google factor, meta description dead last - The September 2026 practitioner survey placing backlinks second.
- Google court ruling reveals quality signals derived from webpage content - What the 2025 antitrust memorandum said about PageRank's role.
- Explaining manual actions - Human-applied penalties, including those for unnatural inbound and outbound links.
- Google clarifies site reputation abuse policy with expanded guidelines - FAQs on relocating third-party content and linking between sites.
- Explaining affiliate - Commission tracking, disclosure duties and the FTC's revised endorsement guides.
- 404 errors don't hurt rankings, says Google Search Relations Team - Martin Splitt on why nofollow and noindex do not signal low quality.
- Google confirms comment spam can't hurt your rankings - John Mueller's January 2026 remarks on spam comment links.
- What Google actually looks for when it catches spam sites - Pedro Dias on manual review and silent PageRank cuts.
- Google's new guide for AI search: what SEO really needs now - Google's 2026 guidance, including nofollow on user-posted links.
- Google spam policies now officially cover AI Overviews and AI Mode in Search - The May 2026 extension of link spam and other policies to generative surfaces.
- Google drops live prices and trip dates in EU Search, then blames Brussels - Context for the August 2026 change to site reputation enforcement in the EEA.
- Sites breaking Google spam rules face lower rankings in two-week window - The September 2026 spam update and the permanence of lost link value.
- Google says new pages take about 20 hours to be found, some never are - Timing data from Barcelona, including link annotations.
Summary
Who: Publishers and platforms apply nofollow, often automatically through content management systems such as WordPress. Advertisers, agencies and affiliates negotiate whether paid placements carry it. Google, Bing and other search engines read it, each on its own terms.
What: A rel attribute value on links, and a robots meta value for whole pages, signalling that a link is not an endorsement and should not pass ranking credit. Google now treats it, together with sponsored and ugc, as a hint rather than a directive.
When: Introduced by Google on January 18, 2005; extended to paid-link disclosure from 2007; joined by sponsored and ugc on September 10, 2019; a crawling and indexing hint at Google since March 1, 2020.
Where: In the HTML of individual links, in robots meta tags and in X-Robots-Tag headers, read by crawlers after a page is fetched. It has no force in law and governs nothing outside search systems that choose to honour it.
Why: Link-based ranking made every link a potential vote, which spammers and link sellers exploited. Nofollow gave publishers a way to withhold that vote, and gave Google a way to separate paid and user-posted links from editorial ones without banning advertising on the web.
Discussion