The short list of rel values that matter to search
Of all the values the HTML rel attribute (the link relation) can carry, only a handful change what search engines do: canonical, alternate when paired with hreflang or media, amphtml, icon, and the link qualifiers nofollow, ugc and sponsored, which Google reads as hints. Two values that once carried weight in Google, next/prev and author, no longer do anything there, and everything else is an instruction to the browser.
The attribute itself names the relationship between the current document and whatever an element points at. The specifications call each keyword a "link type" (WHATWG) or a "link relation type" (RFC 8288 and IANA). It is valid on <link>, <a>, <area> and <form>, and each element accepts a different subset of keywords. The value is a set rather than a single word: space-separated keywords, each listed once, in any order, compared case-insensitively. rel="ugc nofollow" and rel="NOFOLLOW ugc" mean the same thing.
WHATWG sorts link types into three families, and the family predicts whether a search engine could plausibly care. Hyperlinks (canonical, alternate, next, prev, author) point at another document, so only they can describe how one URL relates to another, the raw material of duplicate clustering. External resources (icon, manifest, preload, preconnect, dns-prefetch) are fetched or used by the browser. Annotations (nofollow, noopener, noreferrer) create no link of their own and modify how another hyperlink is treated.
Three registries, and why they disagree
Three sources define rel values, and on a few points they contradict each other.
- The WHATWG HTML Standard holds the types browsers implement, plus some others. It defines
canonicalas a hyperlink keyword for thelinkelement; types browsers do not implement are meant to be registered as extensions elsewhere. - The IANA Link Relations registry, governed by RFC 8288 ("Web Linking," October 2017, which obsoleted RFC 5988), records a defining reference for each type.
canonicalpoints to RFC 6596,amphtmlto the AMP HTML specification, andugcandsponsoredto a Google blog post from September 2019. - The microformats "existing rel values" wiki acts as the HTML extensions registry. It still labels
canonicalandamphtml"proposed" and listscanonicalforaandarea, which conflicts with the WHATWG link-only rule.
For search purposes the tiebreaker is the search engine's own documentation, not a registry. Google reads rel="canonical" only in the <head>, so the microformats listing for anchors is a dead end. And a registry citing a vendor's blog post for ugc and sponsored marks them as search-engine conventions; browsers do nothing with them.
Rel values at a glance
Element columns follow WHATWG and MDN; the last column summarizes what Google, and where noted Bing, documents doing with each value.
| Value | On <link> | On <a> / <area> | What search engines do |
|---|---|---|---|
canonical | Yes | No | Google: a hint for choosing the canonical URL, read only from <head> or an HTTP Link header |
alternate + hreflang | Yes | Yes | Google: identifies localized versions; the HTML link must sit in <head> |
alternate + media | Yes | Yes | Google: connects a desktop page to its separate mobile URL |
alternate + RSS/Atom type | Yes | Yes | Marks a syndication feed; no Google documentation ties it to indexing |
amphtml | Yes | Yes (microformats) | Google: pairs a non-AMP page with its AMP version |
next / prev | Yes | Yes | Google: no longer used. Bing: a discovery hint |
nofollow / ugc / sponsored | No | Yes (nofollow also on <form>) | Google: hints about the linked page; such links generally are not followed |
icon | Yes | No | Google: source of the favicon shown in Search results |
author | Yes | Yes | Google: stopped tracking authorship markup in 2014 |
me | Yes | Yes | No search use documented; Mastodon uses it to verify profile links |
noopener / noreferrer | No | Yes (also <form>) | None; browser security and privacy |
preload / preconnect / dns-prefetch | Yes | No | None; browser performance |
manifest | Yes | No | None; installable web apps |
Two patterns in the element columns trip people up. canonical is link-only, so an anchor carrying it is invalid HTML and is not something Google reads as a canonical declaration. The link qualifiers run the other way: they belong on anchors and mean nothing on a <link>.
How the working values are written
canonical
One link element in the <head>, or the HTTP header for files without one, such as PDFs. Google describes it plainly: "Indicating a canonical preference is a hint, not a rule." Use an absolute URL; Google supports relative ones but warns they can cause trouble later. The rule worth repeating here is about collisions: since February 2024 Google's documentation states that a canonical annotation carrying hreflang, lang, media or type is not used for canonicalization. Canonical and alternate stay in separate tags.
<link rel="canonical" href="https://www.example.com/page">
Link: <https://www.example.com/downloads/white-paper.pdf>; rel="canonical"alternate with hreflang
Each localized version lists every version, itself included, plus an optional x-default. HTML, header and sitemap forms are equivalent to Google, but HTML tags count only inside a well-formed <head>.
<link rel="alternate" hreflang="en-gb" href="https://example.com/en-gb">
<link rel="alternate" hreflang="x-default" href="https://example.com/">alternate with media
Sites with separate mobile URLs annotate the desktop page with the mobile alternate, and the mobile page carries a canonical back to desktop.
On the desktop page:
<link rel="canonical" href="https://example.com/">
<link rel="alternate" media="only screen and (max-width: 640px)" href="https://m.example.com/">On the mobile page, only <link rel="canonical" href="https://example.com/">.
amphtml
The non-AMP page points to its AMP counterpart with rel="amphtml", and the AMP page points back with rel="canonical"; a standalone AMP page is its own canonical. Google indexes AMP pages like other pages, and AMP stopped being a Top Stories requirement with the page experience rollout that began in mid-June 2021.
icon
Google builds the favicon in its results from icon, shortcut icon, apple-touch-icon or apple-touch-icon-precomposed, and the tag has to be on the home page.
nofollow, ugc and sponsored
Google's current guidance assigns each a job: sponsored for advertisements or paid placements, ugc for comments and forum posts, and nofollow when neither fits. Values combine in one attribute. Google announced on September 10, 2019 that all three are hints for ranking, and from March 1, 2020 nofollow also became a hint for crawling and indexing, which ended its role as a dependable crawl block. Bing added support for sponsored and ugc around June 30, 2020.
<a href="https://advertiser.example/offer" rel="sponsored">Partner offer</a>
<a href="https://member-site.example/" rel="ugc nofollow">Posted by a member</a>Values Google has retired: next, prev and author
rel="next" and rel="prev" (or its registered synonym previous) were long standard advice for paginated series. Google's pagination documentation now says it no longer uses these tags, while noting other search engines may. That became public on March 21, 2019, when John Mueller said Google did not use them at all and indicated the change had happened years earlier. The same day, Bing's Frédéric Dubut said Bing reads them as hints for page discovery and understanding site structure, while not merging pages in its index on the strength of them.
So the markup is inert in Google, possibly useful in Bing, and harmless everywhere. The pagination mistake that does cause damage is different: pointing every page's canonical at page 1, which Google's pagination guidance rules out.
rel="author" went the same way earlier. On August 28, 2014, Mueller said Google would no longer track data from content using rel=author markup, ending the authorship experiment. Author links left in templates do nothing for Google and need no cleanup project.
Browser-only values often mistaken for SEO signals
noopenerandnoreferrerare security and privacy controls.noopenerstops the new page from reaching back to the opening window, andtarget="_blank"on<a>,<area>and<form>now implies that behavior anyway.noreferreralso drops the Referer header and impliesnoopener. Neither isnofollow: an anchor markedrel="noopener noreferrer"is an ordinary link as far as search qualifiers go.preload,preconnectanddns-prefetchare performance hints. No search-engine documentation ties them to crawling or indexing.mematters to Mastodon, which verifies a profile link only when the linked page carriesrel="me"in the HTML response itself, readable without running JavaScript. Not every consumer of rel values renders pages.alternatewith an RSS or Atomtypeadvertises a feed. No Google source describes using it as an indexing or canonical signal.
Mistakes that make a rel value invisible
Most rel errors fail silently: the page renders and the search engine simply disregards the annotation.
rel="canonical"on an anchor. WHATWG permits it only onlink, and Google reads it only from the<head>.- Merging canonical and alternate into one tag. A tag such as
<link rel="canonical" hreflang="fr" href="…">is not used for canonicalization under Google's 2024 wording. - Comma-separated keywords. Google tolerates
rel="nofollow, ugc", but HTML defines the list as space-separated, and other parsers owe it no leniency. - An invalid element early in
<head>. Google assumes the head has ended at the first invalid element, such as aniframeorimg, and stops reading. That page does not mention canonicals, but the inference is direct: a link placed after that element is effectively in the body. - Treating
nofollowas a crawl block. Since March 2020 it is a hint; keeping a URL out of the index is a job for other controls.
To find these at scale, a crawler such as Screaming Frog exposes filters on its Canonicals tab named "Outside <head>," "Invalid Attribute In Annotation" and "Multiple Conflicting." For a single URL, Search Console's URL Inspection shows the user-declared canonical beside the Google-selected one.
What to change and what to leave alone
Engineering time on rel values pays off in a narrow band: one canonical in a valid <head>, free of alternate attributes, and correct qualifiers on paid and user-generated links. Much of the rest is tidying with no search consequence.
- Leave existing
next/prevmarkup in place. Removing it gains nothing in Google and may cost Bing a discovery hint. - Do not relabel old
nofollowlinks assponsoredorugcfor neatness. Google presentssponsoredas the preferred label for paid links; any ranking or crawling difference between the three beyond that is unverified. - Do not audit resource hints as SEO items.
preconnectandpreloadbelong in a performance review.
One gap is worth knowing: no current primary source describes how Bing handles hreflang, so treat confident claims about Bing and alternate annotations with skepticism.
Frequently asked questions
Which rel attribute values affect SEO?
In Google, canonical, alternate with hreflang or media, amphtml, icon for the Search favicon, and the link qualifiers nofollow, ugc and sponsored. All of the URL-level ones are signals Google weighs rather than commands it must follow. Values such as noopener, preload and manifest only affect browsers.
Can rel="canonical" be used on an a tag?
No. The WHATWG HTML Standard allows the canonical keyword only on the link element, and Google accepts it only in the <head> or an HTTP Link header. The microformats wiki lists it for anchors, but that listing has no bearing on how Google reads the page.
Does Google still use rel=next and rel=prev?
No. Google's pagination documentation says it no longer uses these tags, and John Mueller confirmed in March 2019 that Google had stopped years before. Bing said it still reads them as discovery hints, so there is no reason to strip existing markup.
What is the difference between nofollow, ugc and sponsored?
Google asks for sponsored on ads and paid placements, ugc on links in comments and forum posts, and nofollow when neither applies. All three are hints, and they can be combined, as in rel="ugc nofollow". Beyond Google's stated preference for sponsored on paid links, no practical difference between them is documented.
Do rel="noopener" and rel="noreferrer" affect SEO?
No search engine documents any effect. noopener is a browser security control, and noreferrer additionally withholds the Referer header. Neither one qualifies a link the way nofollow does, so adding them does not change how the link is treated for search.
Can a rel attribute contain more than one value?
Yes. The value is a set of unique keywords separated by spaces, in any order and in any letter case, so rel="ugc nofollow" is valid. Use spaces, not commas: Google tolerates commas, but HTML does not define them as separators.
Does rel=author still do anything?
Not in Google. In August 2014 John Mueller announced that Google would stop tracking data from rel=author markup. The links are harmless if they remain in templates, but they carry no authorship signal in Google Search.