Product

Solutions

More

Hreflang Audit Checklist: Return Tags, Canonicals and Common Errors

A hreflang audit checklist: build an equivalence table, test return tags and codes, resolve canonical conflicts, and run the audit at scale.

A four-step hreflang audit: build the equivalence table, inspect delivered annotations, verify return links and canonicals, and check the translated content.

A hreflang audit checks that each language or regional version of a page names every other version correctly, names itself, points only at live canonical URLs, and agrees with the page's canonical tag. Start with a table of which pages are equivalents, then compare it with the hreflang annotations your site actually sends, pair by pair. Valid-looking tags can still link the wrong pages.

We make PolyDraft, which generates hreflang for multilingual sites. This checklist works on any site built with any tool, including ours. It's for teams auditing a site that's already localized. If you're still choosing a URL structure or translating the first version, start with our international SEO checklist.

Key points

  • Build the equivalence table before you look at any tags. The audit compares what you intended with what the site sends.
  • Every version lists every other version and itself, with absolute URLs that return 200 and aren't redirected or noindexed.
  • Codes are language first (de, en-GB), never a country alone. en-UK is wrong.
  • A translation's canonical points to itself, never to a page in another language.
  • Search Console no longer reports hreflang errors, so you need a crawler or a script for anything beyond a few pages.

Why a hreflang audit is still needed

Mistakes are common, and Google no longer tells you about them. Ahrefs' 2023 study of 374,756 domains found that 67% had at least one hreflang issue. Search Console's International Targeting report, which used to list hreflang errors, was removed on September 22, 2022. Google still supports hreflang; it just doesn't report problems centrally anymore.

hreflang is also a hint, not a directive. Google can index one version, show another, or group near-identical versions under one canonical. An audit can't force a result. It removes the mistakes that stop Google from understanding which page belongs to which audience.

Build the equivalence table first

Export the intended localized URLs from your CMS or content inventory. Give each group a stable content key, such as product-overview or installation-guide. The key stands for the reader's task, not a translated slug that might change.

Here is a fictional product-page set:

Content key Language or region Expected URL
product-overview English, United States https://example.com/en-us/product
product-overview English, United Kingdom https://example.com/en-gb/product
product-overview German https://example.com/de/produkt

Before checking any tags, open all three pages. They should serve the same purpose, even if prices, terminology or examples differ by market. A German homepage isn't a substitute for a missing German product page. If a translation doesn't exist, leave that version out of the set instead of pointing at a page that only exists in another language.

Three pages pinned to a board and tied to each other: the US English, UK English and German product pages form a reciprocal alternate set, each also listing itself.

Inspect the annotations actually delivered

Google reads hreflang from three places: <link> elements in the HTML head, HTTP Link headers, and XML sitemaps. Pick the one your platform can keep in sync and check that source. Whatever the method, each entry needs a fully qualified URL, each page lists itself, and every pair links both ways.

For the fictional set, this block belongs in the head of all three pages:

<link rel="alternate" hreflang="en-US" href="https://example.com/en-us/product">
<link rel="alternate" hreflang="en-GB" href="https://example.com/en-gb/product">
<link rel="alternate" hreflang="de" href="https://example.com/de/produkt">

In a sitemap, each language version gets its own <url> entry, and each entry lists the full set, including itself. The <urlset> element must also declare xmlns:xhtml="http://www.w3.org/1999/xhtml":

<url>
  <loc>https://example.com/en-us/product</loc>
  <xhtml:link rel="alternate" hreflang="en-US" href="https://example.com/en-us/product"/>
  <xhtml:link rel="alternate" hreflang="en-GB" href="https://example.com/en-gb/product"/>
  <xhtml:link rel="alternate" hreflang="de" href="https://example.com/de/produkt"/>
</url>

For files that have no HTML head, such as PDFs, use an HTTP header:

Link: <https://example.com/en-us/guide.pdf>; rel="alternate"; hreflang="en-US",
      <https://example.com/de/leitfaden.pdf>; rel="alternate"; hreflang="de"

Compare the live output with your table, not the CMS field. A template bug can render a correct saved setting incorrectly. Fetch the raw HTML with curl or "view source", not the browser's inspector: tags added by JavaScript after the page loads may not be read by every crawler. If your site uses more than one method, check that they agree.

Test return tags in both directions

For each pair, record that A points to B and B points to A. A one-way link isn't a complete relationship. In a three-page set, confirm each page's own entry and the two other entries.

Don't stop at counting three tags. Three tags can all point to the same URL, repeat a language value, or name unrelated pages. Compare each code and destination against the expected table.

Validate the values

Most hreflang code mistakes are easy to catch with a short list:

  • Language first, as an ISO 639-1 code: de, ja, pt.
  • Region second and optional, as an ISO 3166-1 alpha-2 code: de-AT, pt-BR.
  • Script where it matters, as an ISO 15924 code: zh-Hans for Simplified Chinese, zh-Hant for Traditional.
  • No country on its own. hreflang="uk" means Ukrainian, not the United Kingdom. The UK is en-GB, never en-UK.
  • Each value once per set, and x-default at most once. This is our advice for keeping sets readable, not a stated Google rule.
  • Absolute URLs that return 200 directly: no redirects, no noindex, no relative paths.
  • In the head, in the HTML the server sends. Google asks for the tags in the <head>; tags placed elsewhere may not be read.

x-default is optional. It names the page for visitors whose language doesn't match any version, such as a language selector or your main language. Pointing it at your default-language page is a common, reasonable choice. Leaving it out isn't an error.

Hreflang and canonical: how to check they agree

The canonical tag names the preferred URL among duplicates. hreflang names the language versions of a page. They answer different questions, so check them separately and then check that they agree. A declared canonical is a signal, not a command.

The rule for translations is simple: A translation canonicalizes to itself; only true same-language duplicates, such as parameter URLs or near-identical regional pages, point elsewhere, and never across languages. A German page whose canonical points to the English page tells Google the German page is a duplicate, and its hreflang entry stops working. Google treats localized pages as duplicates only when the main content is left untranslated, so a page with an English body and German navigation can be grouped with the English page anyway.

Same-language regional pages are the exception you need to plan for. If en-US and en-GB pages are nearly identical, Google may pick one of them as the canonical for both. According to Google's John Mueller, as reported by Search Engine Journal, hreflang can still swap in the right regional URL in results, but Search Console reports the data under the canonical. If each market needs its own data, give each regional page content that's genuinely different: prices, delivery terms, examples.

A stamp choosing one copy from a pile of duplicates beside three pinned pages tied together: canonical picks the preferred page, hreflang lists the language and regional alternatives.

In Search Console's Page indexing report, look for locale URLs under these statuses:

  • Duplicate, Google chose different canonical than user: Google grouped this page with another, often a same-language regional version or an untranslated page.
  • Alternate page with proper canonical tag: expected for parameter or duplicate URLs, but a problem if a real translation appears here.
  • Excluded by 'noindex' tag: a translation listed as an alternate shouldn't be noindexed.

Follow redirects before you record a destination. An alternate that redirects to another locale, a generic homepage or a login page can look correct in the markup while failing the visitor. Record the original URL, the redirect chain and the final page. A slug change without an updated set is the usual cause; our website migration checklist covers redirect maps.

Keep three states apart: a page that's accessible, a page that's declared as an alternate, and a page that's indexed. They aren't the same thing, and a robots.txt block can stop a crawler from reading the page-level tags at all. Our robots.txt examples show what to allow.

Use an expected-versus-actual audit matrix

Sample first, then crawl everything. Start with the homepage, one product or service page and one article for each template, to find systemic bugs quickly. Before a localized release, crawl every language set in full, because individual pages break in ways a template sample can't show: a hand-edited slug, a page published in one language ahead of the others, a page retired in one market only.

Check Expected in the fictional set Evidence to save
Final response Each intended page returns 200 Status and final URL
Locale map Correct en-US, en-GB and de destinations Extracted alternate entries
Return links Each declared pair links both ways Pair-by-pair comparison
Codes Valid, unique values; no country-only codes Extracted values
Canonical Each translation canonicalizes to itself (or a same-language duplicate), never across languages Declared and, where available, Google-selected canonical
Content Same task, right language and market Reviewed headings, body and conversion path
A card file of audit checks with one card pulled up beside a checklist and a magnifier: the illustrative audit flags the German alternate set and asks for canonical and content checks.

Save both the failed output and the corrected output. A later template update can bring a problem back without anyone noticing.

Common hreflang errors and how to fix them

Each failure should point to a place to inspect and a correction you can retest:

Error What you see Likely place to inspect Fix to verify
Missing return link English names German; German doesn't name English Template or a stale locale list Restore the full set on every page
No self-reference A page lists every version except itself Template logic Add the page's own entry
Wrong or country-only code en-UK, uk for the UK, de_DE Locale settings Use language-region codes such as en-GB
Alternate to a redirect or 404 The target's final URL differs or fails Old CMS references after a slug change Point to the final, live URL
Cross-language canonical Every locale page canonicalizes to English Shared SEO template default Make each translation self-canonical
Alternate to the wrong page German product page points to the English homepage Equivalence mapping Point to the equivalent page
Tags present, body untranslated Correct tags, mostly source-language text Translation pipeline Finish and review the main content, or remove the version
Version for a page that doesn't exist An entry for a language the page was never translated into Generated sets that assume every page exists in every language List only versions that exist

These are diagnostic examples, not findings from a specific site.

Run the hreflang audit at scale

Past a few dozen pages, use a crawler or a script, and keep the matrix as the definition of what "correct" means. Tools find mechanical faults; they can't tell you whether two pages are really equivalent.

Matrix row Screaming Frog SEO Spider (Hreflang tab) What the tool can't tell you
Return links Missing Return Links Whether the pair is the right content
Codes and regions Incorrect Language & Region Codes Whether that market version is the one you want shown
Canonical Non-Canonical Return Links Which canonical Google actually chose (use URL Inspection)
Indexing Noindex Return Links Whether the page is indexed
Coverage Unlinked Hreflang URLs (needs Crawl Analysis) Whether a missing translation should exist

Filter names are from Screaming Frog's documentation as of November 17, 2026; they may change. Site-audit tools from Ahrefs and Semrush include similar hreflang checks under their own names.

For sites on several domains (example.de, example.fr), each domain's pages must list the others, and the crawler must be allowed to fetch the other domains or it can't confirm the return links on those domains.

For a quick check without a crawler, a short script can fetch a set of URLs, read the alternates from the HTML each one sends, and build the pair matrix. We wrote one for this article and ran it on November 6, 2026 against a three-page test set with four deliberate faults (an en-UK code, a cross-language canonical, an alternate pointing at a missing page, and a missing return link). It flagged all four. Run on polydraft.ai, which is in one language, it correctly reported no alternates. Whatever you use, confirm a sample of its findings by hand.

Search Console's URL Inspection then shows, for a few important URLs, which canonical Google selected. It doesn't replace the matrix, and a live test doesn't guarantee which regional page Google shows for every search.

Verify language content and the full reader journey

Read each translated page as a visitor. Check the title, headings, body, navigation, image descriptions, form labels and confirmation message. Check that the call to action leads somewhere usable in that language. A polished product page followed by an untranslated form is an unfinished localization.

Google recommends separate URLs for each language version. Don't rely only on redirecting visitors by IP address or browser language; that can stop people and crawlers from reaching a specific version. Offer a language switcher that links to the equivalent page, not the homepage.

After a fix:

  1. Save the intended locale set and the previous configuration.
  2. Apply the mapping or template correction.
  3. Fetch every affected page and annotation source again.
  4. Repeat the canonical, return-link and content checks.
  5. Check language navigation and the conversion path.
  6. Use URL Inspection for a few important URLs and record the result separately.

Re-run the audit after any redesign or move; our website redesign SEO checklist lists it as a release check. If your locale URL structure itself is still in question, website architecture for SEO covers how to set URL rules once.

How PolyDraft generates hreflang, and what to check anyway

On new sites built with PolyDraft's current template, the hreflang tags, the x-default entry and the sitemap alternates are generated, not typed by hand:

  • The sitemap uses the xhtml:link format, and blog posts are grouped by their actual translations.
  • URL paths use lowercase locale codes (/es-mx/), while hreflang values use standard casing (es-MX).
  • x-default points to the site's default language.
  • A separate-domain fork is set up to list its sibling sites; check that the original site links back.
  • The localization workflow ends with an audit step the agent runs, covering page parity and hreflang completeness. It's a workflow check, not a guarantee.

Limits: you build the primary language first and add languages one at a time, up to the number your plan includes (see language limits by plan). You can review each language before publishing. Generated tags still need this audit. One limit to check first: today, when a page or blog post has no translation in a language, the page's <head> can still list that language as an alternate. That URL can redirect (static pages) or show the default-language content without a noindex (blog posts). The sitemap leaves untranslated blog posts out but lists static pages in every language, so the <head> and the sitemap can disagree. Run the matrix on those pages first, exactly as you would on any other site. Platforms that generate tags for you, such as Webflow's Localize or Wix Multilingual, need the same check.

Frequently asked questions

Do I need hreflang on a single-language website?

No. hreflang only describes language or regional versions that exist. A site in one language has nothing to annotate, and adding entries for versions you don't have creates errors.

Does hreflang translate the page?

No. It tells search engines which existing pages are versions of each other. The body, navigation and forms still have to be translated. Marking an English page as German doesn't make it German.

Should every translation point its canonical to English?

No. Each translated page should canonicalize to itself. A canonical that points to another language tells Google the translation is a duplicate, and its hreflang entry is then ignored.

Is a self-referencing hreflang required?

Yes, include it. Google's guidance says each page lists itself as well as the other versions. Ahrefs' 2023 study found it missing on 18% of the sites it checked; adding it is usually a template change.

Do I need x-default?

Not required: Google calls x-default optional. We recommend it when you have a sensible page for visitors who match none of your versions, such as a language selector or your main-language page. If you add it, use the same x-default URL on every page in the set.

Is it en-UK or en-GB?

en-GB. Region codes are ISO 3166-1, where the United Kingdom is GB. A country code alone isn't valid either: uk means Ukrainian.

Should hreflang point only to canonical URLs that return 200?

Yes. Each entry should be the final, canonical, indexable URL. Redirects, 404s, noindexed pages and non-canonical variants weaken or break the set.

Why did Google choose a different canonical for my regional page?

Usually because it looks like a duplicate: the same language with near-identical content, or a translation whose main content is still in the source language. Make the page genuinely distinct, or accept one canonical and let hreflang swap the URL shown.

Make locale correctness part of every release

Keep the equivalence table with your content inventory, and re-run the affected checks whenever a page is added, removed or renamed. Retire a page in every language in the same change. For the wider localization process, see how PolyDraft handles website localization. A useful audit ends with evidence for each relationship, not a count of tags.

About this article. Drafted by an AI agent following PolyDraft's blog workflow, checked by an automated QA script, reviewed by separate AI agents for facts and editorial quality, and approved by Vincent Chen. First published October 8, 2026; updated November 17, 2026 with sections on codes, canonicals, common errors and tools. Google documentation was checked on November 17, 2026. The examples and diagrams are illustrative, not findings from a customer site. The test script results are from our own run on November 6, 2026. PolyDraft's behavior is described from its current product code as of November 17, 2026. For corrections, contact us with the page URL and a source.

Continue with the next task

Sources

Founder of PolyDraft

Vincent Chen is the founder of PolyDraft and CEO of BlackMonolith, Inc. He has worked on technical SEO for multiple products and websites, and runs polydraft.ai on PolyDraft.

Put your site’s SEO and GEO on autopilot

PolyDraft builds your website, then finds the keywords you can win every week and writes the content.

Start your site

Start with a brief.
Publish when it’s right.

Build and preview your whole site for free. Every week after launch, PolyDraft finds the keywords you can win and writes the content.

Free to build and preview. Pay only to publish.