rel="alternate": What It Does and When to Use It

rel="alternate" is one HTML link relation with several unrelated SEO jobs — hreflang language variants, RSS/Atom feed discovery, and legacy mobile/AMP alternates. Which job you're doing decides which attributes belong next to it, and which of these still matter in 2026.

First published: Jul 2, 2026 · Last updated: Jul 25, 2026 · Advanced
demand #14 in Meta Tags#47 in On-Page#219 in Technical SEO#302 on the site

rel="alternate" is a general-purpose HTML link relation with several unconnected SEO jobs, and the attribute next to it decides which job: hreflang for language/region variants (the one that matters most today), type="application/rss+xml" for RSS/Atom feed autodiscovery, and media for a separate mobile URL (legacy). A close cousin, rel="amphtml", points to an AMP page. Three of these four — hreflang, mobile-alternate, and AMP — need the relationship declared from both ends (reciprocal pairing); RSS/Atom feed autodiscovery is the exception and works one-way, no return link required. None of them is a ranking lever or a guarantee that a page gets crawled, indexed, or consolidated: hreflang routes the right variant to the right searcher, feed links aid discovery, and AMP lost its special treatment in 2021 (Google ended cache-served AMP pages in Search on July 1, 2026). Think of rel="alternate" as a routing/consolidation hint that says "these different URLs are intentional relatives, not accidental duplicates" — not a directive search engines must follow.

TL;DR — rel="alternate" is the WHATWG-defined “alternate” link relation, and its meaning is set entirely by the attribute paired with it: hreflang → language/region variant (current, high-value — full rules in the hreflang guide); type="application/rss+xml" / "application/atom+xml" → feed autodiscovery (current, niche, discovery-only); media → separate-mobile-URL variant (legacy — responsive design is Google’s default now). A related but distinct tag, rel="amphtml", points a canonical page at its AMP version. The throughline is that each alternate relationship has its own syntax and validation rules — three of the four (hreflang, mobile-alternate, AMP) require reciprocal pairing from both ends; RSS/Atom feed autodiscovery does not. None of these is a ranking lever, and none guarantees crawling, indexing, or consolidation; rel="alternate" is fundamentally a routing or discovery hint, not a directive.

Evidence for this claim HTML link rel=alternate indicates an alternate representation, such as another language or feed, with semantics refined by attributes like hreflang and type. Scope: HTML alternate link relation. Confidence: high · Verified: WHATWG HTML: Link type alternate Evidence for this claim Google uses rel=alternate hreflang for localized URL alternates when annotations are valid and reciprocal; it does not consolidate them like rel=canonical. Scope: Current Google hreflang behavior. Confidence: high · Verified: Google Search Central: Localized versions

What rel="alternate" actually is

It’s a value of the HTML link element’s rel (relationship) attribute, defined in the WHATWG HTML Living Standard’s link-type list. On its own, “alternate” just means “an alternate representation of the current document.” The specific meaning comes from the other attribute you pair it with. That’s why one tag ends up doing several unrelated SEO jobs — and why generalist “rel=alternate explained” posts tend to blur them together.

I’ve long framed this tag through the lens of duplicate content. In my Search Engine Land piece The myth of the duplicate content penalty, I listed it as one of the tools that “consolidate alternate versions of a page, such as mobile or various country/language pages.” That’s the unifying thesis worth holding onto: every use of rel="alternate" exists to tell search engines these different URLs are intentional relatives, handle them as one entity — the opposite of accidental duplication. (For how engines actually pick and consolidate a representative URL, see the sibling topic on canonicalization.) None of these annotations guarantees an outcome — not crawling, not indexing, not consolidation, not rankings or traffic, not feed subscriptions, and not citation by an AI answer engine. They’re hints search engines can act on, not directives they must follow.

The four SEO uses (and one close cousin)

1. Hreflang — language/region alternates (current, most important)

This is the one that matters most in 2026. Paired with hreflang, rel="alternate" declares a language or regional variant:

<link rel="alternate" hreflang="es" href="https://example.com/es/page/" />

Google’s framing: “If you have multiple versions of a page for different languages or regions, tell Google about these different variations,” using hreflang “so that we can understand that these pages are localized variations of the same content.” The load-bearing rule is reciprocity — Google: “Each language version must list itself as well as all other language versions.” Break that and Google ignores the pair (it’s a hint, not a directive — wrong hreflang is ignored, not penalized).

That’s the mental model. I’m deliberately not re-deriving ISO language/region codes, x-default, or the three implementation methods here — that all lives in the dedicated hreflang deep dive on this site. Go there for implementation; stay here for the mechanism-level view.

2. RSS/Atom feed autodiscovery (current, niche)

Paired with a feed type, rel="alternate" tells browsers and feed readers where your syndication feed lives:

<link rel="alternate" type="application/rss+xml" title="Example Feed"
      href="https://example.com/feed.xml" />

Swap application/atom+xml for an Atom feed. This is the use almost every generalist article forgets, even though “rel alternate rss” is a real query cluster. What it does not do is boost rankings — John Mueller has been explicit that “if you are looking for a ranking boost by having an RSS feed, that’s not going to happen” and that “there’s no direct ranking boost for the website itself” (Search Engine Roundtable coverage). The real value is discovery: feeds work like a lightweight sitemap, helping Google find new and updated URLs faster (Lumar’s office-hours summary of Mueller’s comments). Discovery, not ranking.

3. Mobile-alternate (separate mobile URLs) — legacy

Paired with media, rel="alternate" points a desktop URL at a separate mobile (m-dot) URL:

<link rel="alternate" media="only screen and (max-width: 640px)"
      href="https://m.example.com/" />

…reciprocated by a rel="canonical" on the mobile page pointing back to the desktop URL. This still appears in Google’s mobile-first indexing best-practices docs (updated December 2025), but it’s explicitly the non-preferred option: Google “recommends Responsive Web Design because it’s the easiest design pattern to implement and maintain,” and files the separate-URLs annotation under “additional best practices” — i.e., legacy support. With mobile-first indexing complete for all sites since July 2024, Google evaluates the mobile version of a page regardless of your setup. Don’t build new mobile-alternate setups in 2026; only maintain one if you’re still stuck on m-dot URLs and can’t migrate to responsive yet.

Here’s a disambiguation almost nobody makes cleanly: rel="amphtml" is not rel="alternate". It’s a different rel value in the same “here’s another version of this page” family, and the two get conflated constantly. It points a canonical non-AMP page at its AMP version:

<link rel="amphtml" href="https://example.com/page.amp.html" />

…reciprocated by rel="canonical" on the AMP page pointing back. The markup still works exactly as documented for any site that keeps AMP, but AMP-the-strategy lost its reason to exist years ago (see the 2026-status section below). They’re cousins, not the same tag.

Which of these still matter in 2026

UseAttribute pairingStatus in 2026
HreflanghreflangCurrent, high-value — the one that matters
RSS/Atom feedtype="application/rss+xml"Current, niche — discovery only, no ranking effect
Mobile-alternatemediaLegacy — maintain, don’t newly adopt
AMP (rel="amphtml")(different rel value)Optional since 2021; infra wound down

AMP specifically: Google removed the requirement that pages be AMP to appear in Top Stories in 2021 and retired the lightning-bolt badge that July, making Core Web Vitals / page experience the operative signal instead of the AMP format itself (Google’s page-experience update post; Search Engine Land). Google now “indexes AMP pages just like other web pages, and applies the same standard to all pages, regardless of the technology used.” Then on July 1, 2026, Google ended cache-served AMP pages in Search entirely — clicks now route straight to the publisher’s own AMP-hosted page instead of a google.com AMP viewer URL, and Google removed references to “the AMP viewer, AMP Cache, and signed exchanges” from its docs (Search Engine Journal; Search Engine Roundtable). If you still run AMP, keep the rel="amphtml" markup correct; nobody should be adopting AMP anew for SEO reasons.

The throughline: reciprocal pairing

Three of the four uses want the relationship declared from both ends — RSS/Atom feed autodiscovery is the exception; it’s a one-way discovery link with no return-tag requirement, and Google’s own hreflang docs scope the reciprocity rule to localized-version clusters, not to every rel="alternate" use:

  • hreflang — every version lists itself and every other version.
  • mobilerel="alternate" on desktop, rel="canonical" back on mobile.
  • AMPrel="amphtml" on the canonical page, rel="canonical" back on the AMP page.

Broken reciprocity is the single most common implementation bug across these three. If a relationship that’s supposed to be reciprocal is one-way, expect it to be ignored or half-declared — RSS/Atom needs no return link at all.

Each alternate system has its own return signal; RSS and Atom are discovery links, not reciprocal page pairs. Source: Patrick Stox

For hreflang, the English and Spanish pages list themselves and one another. For legacy separate mobile URLs, the desktop page uses rel alternate with media and the mobile page points back with rel canonical. AMP is a related but distinct pattern: the canonical page uses rel amphtml and the AMP page points back with rel canonical. RSS and Atom feed autodiscovery is one-way and is not a reciprocal page pairing.

© Patrick Stox LLC · CC BY 4.0 ·

Common mistakes

  • Missing the reciprocal link. The number-one bug. A one-way rel="alternate" relationship usually gets ignored (hreflang) or leaves the pairing half-declared (mobile/AMP).
  • Confusing rel="alternate" with rel="amphtml". Different rel values, different jobs. AMP uses rel="amphtml", not rel="alternate".
  • Expecting a ranking boost. None of these lift positions. hreflang routes; feed links aid discovery; AMP hasn’t carried ranking weight since 2021.
  • Building new mobile-alternate or AMP setups in 2026. Responsive design is the default; AMP has no SEO upside. These are maintenance patterns, not build patterns.
  • Running separate mobile URLs and responsive design at once. Mueller has warned this can confuse Google’s processing (Search Engine Roundtable) — pick one.

Bing and other engines

Bing supports hreflang via rel="alternate" but treats it as comparatively weak. Microsoft’s Fabrice Canel has said “hreflang is indeed a far weaker signal than content-language at Bing” (via Search Engine Roundtable) — so at Bing, rel="alternate"/hreflang is one input among several (the content-language header/meta, <html lang>, inbound links, and visitor geography) rather than authoritative. Full Bing/hreflang treatment lives in the hreflang guide.

Where this sits

rel="alternate" is a link-level signal that lives in your <head> alongside the rest of the meta tags that matter for SEO — it’s a sibling of the robots meta tag, the title element, and the meta description in that on-page cluster, even though it’s technically a link element rather than a meta one. Its closest conceptual relative is canonicalization: canonical says “these duplicates are the same page, pick one,” while rel="alternate" says “these are intentionally different versions, keep them all but understand they’re related.” For the hreflang-specific rules — codes, x-default, the three implementation methods, auditing clusters at scale — see the dedicated hreflang guide.

Add an expert note

Pin an expert quote

New person? Create their unclaimed profile at /admin/experts/ → Pin an expert quote first.