---
title: "hreflang errors: what Polish homepages get wrong"
description: "One in six multilingual .pl homepages gets hreflang, canonical or lang wrong, most often a missing self-reference. Codes, x-default and the fixes."
canonical_url: "https://www.outofplace.space/blog/hreflang-polish-web"
language: "en"
last_updated: "2026-11-17"
alternates:
  pl: "https://www.outofplace.space/pl/blog/hreflang-wersje-jezykowe-polskich-stron.md"
---

# hreflang errors: what Polish homepages get wrong

> One in six multilingual .pl homepages gets hreflang, canonical or lang wrong, most often a missing self-reference. Codes, x-default and the fixes.

Published 2026-11-17 · Category: Search & AI · Tags: hreflang, International SEO, Poland, CMS

Polish websites with language versions mostly mark them with hreflang, and most of the hreflang errors they make come down to one omission. Of the 7,415 popular .pl homepages we read, 1,151 use hreflang to point to versions in other languages. 17% of those have at least one error on the homepage itself, by Google's hreflang rules or by our two checks of the canonical and `<html lang>`, and the most common error by far is a page that lists its English or German version but not itself: 12% of them. Codes that aren't valid, such as `ua` for Ukrainian and `cz` for Czech, turn up on only 4.1%.

We read the HTML and the response headers of every .pl homepage in [the Tranco list of popular sites](https://tranco-list.eu/list/Y8YYG/full) that answered on 30 October 2026, the same sample as our study of [WordPress market share in Poland and the other CMSs of .pl sites](https://www.outofplace.space/blog/wordpress-polish-web), and checked each hreflang set against [Google's rules for localized versions](https://developers.google.com/search/docs/specialty/international/localized-versions). The homepage is all we fetched, so this is what a homepage shows: whether the other versions answer back needs those pages too.

- 16% of .pl homepages mark language versions with hreflang. Of homepages with any sign of language versions, 74% do.
- 12% of multilingual homepages don't list themselves in their own hreflang set, the most common error by a wide margin.
- About half have no x-default, which Google recommends but doesn't require. Only 40% pass every check and have one.
- On Shopify, IdoSell, Magento and PrestaShop, at least nine in ten homepages with language versions mark them with hreflang; homepages with no CMS we could recognise, about three in five.
- 122 homepages carry hreflang that names no language besides their own and points at another address: all 12 we checked by hand pointed at gambling pages.

## How do Polish websites mark their language versions?

With hreflang link elements in the `<head>`, almost without exception. An annotation is one `<link rel="alternate" hreflang="en" href="…">` per language version, and 1,426 homepages (19%) carry at least one. We count a homepage as multilingual when its set, together with the page's own language, names two languages or more; that leaves out the homepages whose hreflang names no language but their own, which we come back to below.

Google accepts the same annotations in an HTTP header or an XML sitemap. Headers are rare: 14 multilingual homepages send a `Link` header with hreflang, 4 of them without the HTML tags. We don't read sitemaps, so a site that marks its versions only there counts here as having none.

Nine in ten multilingual homepages name English, and almost half German. Ukrainian is in 18% of the sets, close to Russian.

**Polish multilingual homepages point to English first, then German** (Share of .pl homepages with language versions whose hreflang set names each language (Polish aside) · 95% intervals)

| Item | Value | 95% interval | n |
| --- | --- | --- | --- |
| English | 91% | 89%–92% | 1,151 |
| German | 45% | 43%–48% | 1,151 |
| Czech | 29% | 27%–32% | 1,151 |
| French | 28% | 25%–30% | 1,151 |
| Spanish | 27% | 25%–30% | 1,151 |
| Italian | 25% | 22%–27% | 1,151 |
| Dutch | 21% | 19%–23% | 1,151 |
| Slovak | 21% | 18%–23% | 1,151 |
| Romanian | 20% | 18%–22% | 1,151 |
| Ukrainian | 18% | 16%–20% | 1,151 |
| Hungarian | 17% | 15%–19% | 1,151 |
| Portuguese | 17% | 15%–19% | 1,151 |
| Russian | 15% | 13%–17% | 1,151 |

Source: Outofplace census, 30 Oct 2026; Tranco list Y8YYG. Chart: https://www.outofplace.space/data/blog/hreflang-polish-web/charts/en/chart-polish-multilingual-homepages-point-to-english-first-then-german.png Data: https://www.outofplace.space/data/blog/hreflang-polish-web/2026-10-30/languages.csv

### Is hreflang needed for a multilingual site?

Google asks for it when you have versions of a page in different languages or for different regions, so its results can show each user the right one. Most Polish sites with versions do have it. The rule that finds language-switch links (a link to `/en/`, `en.` or `?lang=en` on the same site) found 406 homepages that link to another language version and mark none with hreflang; all 12 we checked by hand were real language versions. Another 132 homepages without hreflang load a translation widget that translates the page in the browser, with no separate address for Google to show.

It varies by platform. On Shopify, IdoSell, Magento and PrestaShop, at least nine in ten homepages with language versions mark them with hreflang. Homepages with no CMS we could recognise do it on about three in five, level within the margin with Drupal and the smaller CMSs.

**Shop platforms mark language versions with hreflang almost always** (Share of .pl homepages with language versions (hreflang for two languages, or a link to a language version) whose hreflang marks them, by CMS or platform · 95% intervals)

| Item | Value | 95% interval | n |
| --- | --- | --- | --- |
| Shopify | 100% | 93%–100% | 54 |
| IdoSell | 98% | 91%–100% | 60 |
| Magento | 95% | 83%–99% | 39 |
| PrestaShop | 91% | 86%–95% | 164 |
| Other shop platforms | 88% | 73%–95% | 34 |
| JS framework, no CMS seen | 87% | 80%–92% | 123 |
| WordPress | 77% | 72%–81% | 357 |
| Other CMSs and platforms | 68% | 58%–76% | 99 |
| Drupal | 62% | 48%–75% | 45 |
| No CMS detected | 59% | 55%–63% | 582 |

Source: Outofplace census, 30 Oct 2026; CMS groups from our WordPress study. Chart: https://www.outofplace.space/data/blog/hreflang-polish-web/charts/en/chart-shop-platforms-mark-language-versions-with-hreflang-almost-alway.png Data: https://www.outofplace.space/data/blog/hreflang-polish-web/2026-10-30/checks.csv

On WordPress the share differs by translation plugin. Every homepage we saw that loads assets of Polylang, TranslatePress, qTranslate or MultilingualPress marks two languages or more with hreflang, though that group is small. WPML homepages do so 60% of the time, and those running an in-browser translation widget such as GTranslate 23%.

**On WordPress, every site we saw with Polylang, TranslatePress or a similar plugin had hreflang** (Share of WordPress homepages with hreflang for two languages or more, by the translation plugin whose assets the homepage loads · 95% intervals)

| Item | Value | 95% interval | n |
| --- | --- | --- | --- |
| No translation plugin seen | 11% | 9.3%–12% | 1,477 |
| Polylang, TranslatePress, other | 100% | 91%–100% | 37 |
| Translation widget (GTranslate and others) | 23% | 13%–36% | 53 |
| WPML | 60% | 51%–68% | 112 |

Source: Outofplace census, 30 Oct 2026; plugins from our WordPress study. Chart: https://www.outofplace.space/data/blog/hreflang-polish-web/charts/en/chart-on-wordpress-every-site-we-saw-with-polylang-translatepress-or-a.png Data: https://www.outofplace.space/data/blog/hreflang-polish-web/2026-10-30/checks.csv

## What are the most common hreflang errors?

We checked each multilingual homepage's set against the rules Google states for the method: every version lists itself and every other version, codes are ISO 639-1 languages with an optional ISO 3166-1 alpha-2 region, URLs are fully qualified, and the tags sit in the `<head>`. We added two checks of our own: a canonical that agrees with the page, and an `<html lang>` that agrees with the page's own code.

**A missing self-reference is the most common hreflang error** (Share of .pl homepages with language versions marked by hreflang whose own set fails each check · 95% intervals)

| Item | Value | 95% interval | n |
| --- | --- | --- | --- |
| Any of the errors below | 17% | 15%–19% | 1,151 |
| Doesn't list itself | 12% | 10%–14% | 1,151 |
| A code that isn't valid | 4.1% | 3.1%–5.4% | 1,151 |
| Canonical points away from the page | 3.8% | 2.9%–5.1% | 1,151 |
| A relative URL | 1.6% | 1%–2.5% | 1,151 |
| One code, two addresses | 1% | 0.5%–1.7% | 1,151 |
| html lang names another language | 0.4% | 0.2%–1% | 1,151 |

Source: Outofplace census, 30 Oct 2026. Chart: https://www.outofplace.space/data/blog/hreflang-polish-web/charts/en/chart-a-missing-self-reference-is-the-most-common-hreflang-error.png Data: https://www.outofplace.space/data/blog/hreflang-polish-web/2026-10-30/checks.csv

The missing self-reference is about three times as common as the next error. Every other error is on fewer than one homepage in twenty, and the `<html lang>` almost always agrees with the code the page gives itself: 5 homepages disagree. No multilingual homepage put its tags in the `<body>`, and none wrote a code with an underscore (`en_GB`), the locale format many programming languages use.

### Does each page need a self-referencing hreflang tag?

Yes. Google's documentation says that "each language version must list itself as well as all other language versions", and that "if two pages don't both point to each other, the tags will be ignored". A Polish homepage that lists only `en` and `de` gives the English page nothing to point back to: the English page can list the Polish one, but the Polish one never confirms it.

138 multilingual homepages fall short of that in one of three ways. 98 list no URL of their own at all. 19 name themselves only as x-default, so Google is told this is the fallback, not the Polish version. And 21 list themselves at an address that redirects to the homepage or differs from it by `www`, the protocol or a trailing slash. That last group looks right in the source and still points at a different URL.

### Which hreflang codes are wrong?

Few, and they are predictable. 47 multilingual homepages use at least one code that isn't valid: 26 put a country code where the language goes, and 18 add a region that isn't a country in ISO 3166-1. For Polish sites the two typical ones are the neighbours: `cz` for Czech, which is `cs`, and `ua` for Ukrainian, which is `uk`. Against 205 homepages that mark Ukrainian correctly, 11 use `ua` or a `ua-` code. `en-uk` (the United Kingdom is `gb`) turned up once.

Most valid sets use the bare language. Of the homepages that list themselves, 721 give themselves a language alone (`pl`), 263 add the country (`pl-pl`) and 29 carry both. Both forms are valid. The country only matters when you have versions of one language for different countries, such as `de-de` and `de-at`.

### Canonical and hreflang together

[Google's guide to canonical URLs](https://developers.google.com/search/docs/crawling-indexing/consolidate-duplicate-urls) asks you to "specify a canonical page in the same language, or the best possible substitute language", and says it prefers URLs that are part of hreflang clusters. 44 multilingual homepages (3.8%) contradict their own set: 19 name another language version as canonical, which tells Google the Polish page is a copy of it, and the rest point it at another address or a variant of their own. 12% have no canonical, which isn't an error.

### Do errors depend on the CMS?

They differ, though the margins of the smaller groups are wide. Homepages with no CMS we could recognise have an error on more than one in four, level within the margin with the smaller shop platforms, other CMSs and frameworks. On WordPress, PrestaShop, Shopify and IdoSell it is about one in ten. How the same shop platforms compare on speed is in our [Core Web Vitals study of Polish online shops](https://www.outofplace.space/blog/core-web-vitals-polish-ecommerce).

**On WordPress, PrestaShop, Shopify and IdoSell about one in ten hreflang sets has an error** (Share of .pl homepages with language versions marked by hreflang whose own set fails at least one check, by CMS or platform · 95% intervals)

| Item | Value | 95% interval | n |
| --- | --- | --- | --- |
| No CMS detected | 28% | 23%–33% | 344 |
| Other shop platforms | 23% | 12%–41% | 30 |
| Other CMSs and platforms | 21% | 14%–30% | 95 |
| JS framework, no CMS seen | 18% | 12%–26% | 107 |
| Magento | 14% | 5.9%–28% | 37 |
| IdoSell | 10% | 4.7%–20% | 59 |
| Shopify | 9.3% | 4%–20% | 54 |
| WordPress | 9.1% | 6.2%–13% | 275 |
| PrestaShop | 8.7% | 5.1%–14% | 150 |

Source: Outofplace census, 30 Oct 2026; CMS groups from our WordPress study. Chart: https://www.outofplace.space/data/blog/hreflang-polish-web/charts/en/chart-on-wordpress-prestashop-shopify-and-idosell-about-one-in-ten-hre.png Data: https://www.outofplace.space/data/blog/hreflang-polish-web/2026-10-30/checks.csv

## What is x-default and when should you use it?

`x-default` is the reserved value for the page Google shows "when no other language/region matches the user's browser setting": usually a language chooser, or the version you'd rather show a visitor you know nothing about. It is a recommendation, not a rule, and 51% of multilingual Polish homepages don't have one. For a Polish site whose root is the Polish version, the root can be both `pl` and `x-default`; what it shouldn't be is x-default alone.

## Folder, subdomain or separate domain?

Folders win. On 60% of multilingual homepages every other version lives in a folder of the same host (`/en/`, `/de/`), and on 32% on another domain, such as a brand's country domains (`.de`, `.cz`) or a `.com`. Subdomains and a language parameter are rare. Google's [guide to multi-regional sites](https://developers.google.com/search/docs/specialty/international/managing-multi-regional-sites) describes all of them, and hreflang works the same way in each.

**Most Polish sites put their language versions in folders** (Where the other versions in a multilingual .pl homepage's hreflang set live, x-default aside · 95% intervals)

| Item | Value | 95% interval | n |
| --- | --- | --- | --- |
| Folders (/en/) | 60% | 57%–63% | 1,151 |
| Separate domains (.de, .com) | 32% | 29%–34% | 1,151 |
| A mix | 5.4% | 4.2%–6.9% | 1,151 |
| Subdomains (en.) | 1.3% | 0.8%–2.1% | 1,151 |
| A parameter (?lang=en) | 1.2% | 0.7%–2% | 1,151 |

Source: Outofplace census, 30 Oct 2026. Chart: https://www.outofplace.space/data/blog/hreflang-polish-web/charts/en/chart-most-polish-sites-put-their-language-versions-in-folders.png Data: https://www.outofplace.space/data/blog/hreflang-polish-web/2026-10-30/checks.csv

## hreflang that points to gambling pages

122 homepages (1.6% of the sample) carry hreflang that names no language besides their own and points at another address; on 70% of them the canonical points there too. We checked 12 of them by hand, drawn at random, judging each target by its URL and the homepage it sits on, and all 12 pointed at online casino or betting pages: on other domains, on subdomains of the site, or in a folder of it. A sample that small can't tell what every one of these homepages is, but such tags are easy to miss in a page's source. If your homepage's hreflang names a page you don't recognise, look at it as a possible security problem, not only an SEO one.

## Does Google use the HTML lang attribute?

Not for language. Google's guide states that it "doesn't use hreflang or the HTML lang attribute to detect the language of a page"; it reads the content. The attribute still matters to screen readers, browser translation and hyphenation. Of the 6,472 homepages whose text reads as Polish, 6.7% have no `lang` and 2% say they are English.

Bing is the other engine Polish sites serve. Its webmaster blog, in [a post on duplicate content from December 2025](https://blogs.bing.com/webmaster/December-2025/Does-Duplicate-Content-Hurt-SEO-and-AI-Search-Visibility), tells site owners to "use hreflang to define language and regional targeting" and to audit hreflang relationships with canonicals and redirects.

## How to fix hreflang errors

For each language version of a page, in this order:

**1. List every version, the page itself included.**
The set is the same on every version: each one lists itself and all the others, with full `https://` URLs of the canonical address, not one that redirects.

```html
<link rel="canonical" href="https://www.example.pl/" />
<link rel="alternate" hreflang="pl" href="https://www.example.pl/" />
<link rel="alternate" hreflang="en" href="https://www.example.pl/en/" />
<link rel="alternate" hreflang="uk" href="https://www.example.pl/uk/" />
<link rel="alternate" hreflang="x-default" href="https://www.example.pl/" />
```

**2. Use language codes, and countries only when they differ.**
ISO 639-1 for the language (`cs`, `uk`, `de`), then an ISO 3166-1 country only when you have versions per country (`de-at`). Never a country alone (`cz`, `ua`, `at`), never `en-uk` (the United Kingdom is `gb`), never an underscore.

**3. Point the canonical at the page itself.**
Each version is its own canonical. A Polish page whose canonical is the English one tells Google it is a duplicate of it.

**4. Add x-default for visitors no version matches.**
Point it at the language chooser or at the version you'd show anyone else. On a Polish-first site that is often the Polish root, listed as both `pl` and `x-default`.

**5. Set html lang to the page's language.**
`<html lang="pl">` on the Polish version, `lang="en"` on the English one. Google doesn't read it for language, but screen readers and browsers do.

**6. Check the other side.**
Open each version's source and confirm it lists the same set. Search Console no longer reports hreflang errors: Google [deprecated the International Targeting report](https://support.google.com/webmasters/answer/12474899?hl=en) and says it will keep supporting and using hreflang tags. A crawler that follows the alternates is the way to check return links across the site.

## How we measured

The homepages come from our census of 30 October 2026 by `OutofplaceResearchBot` (see [what our crawler fetches and how to block it](https://www.outofplace.space/bot)), which reads robots.txt first and fetches only the homepage: 7,648 of the 9,208 .pl domains in the Tranco list answered with HTML, and 7,415 distinct sites remained after merging domains that redirect to the same address.

We read each homepage's HTML as it arrived, without running scripts, with comments, scripts and styles removed: every `<link rel="alternate">` with an `hreflang`, and the `Link` header of the final response. Each value is checked against ISO 639-1 and ISO 3166-1 alpha-2 (with an optional ISO 15924 script), as Google's rule says. A homepage lists itself when a language code points at its final URL; an address in its redirect chain, or one that differs only by protocol, `www` or a trailing slash, counts as a variant and not as itself. CMS groups and WordPress plugins come from [our WordPress study](https://www.outofplace.space/blog/wordpress-polish-web#what-share-of-polish-sites-run-wordpress) of the same census; groups under 30 homepages are folded into "Other". Shares carry 95% Wilson intervals.

We drew 12 homepages at random (seed 7) from each of two groups and read their tags and link targets by hand. Homepages with links the rules take for language versions but no hreflang for two languages: 12 of 12 were language versions (switchers to `/en`, `en.` subdomains, `?lang=en`). Homepages whose hreflang names no language besides their own and points away: 12 of 12 pointed at gambling pages, judged by the target's URL and the homepage it sits on. The labels are in the dataset. The switch-link rule finds language versions; it is no proof that a site without such links has none, so the share of sites with versions that use hreflang is an upper bound.

We saw the homepage only. Return links need the other versions fetched, so a set that is complete on the homepage can still be broken on the English page; redirects, errors and noindex on the alternate URLs are invisible too. Annotations in XML sitemaps and those a script inserts aren't read, so a site that uses only its sitemap counts as having no hreflang. Homepages over 1 MiB (319) were cut. Nothing here measures rankings or traffic: the CMS and plugin differences are correlations between sites that differ in other ways too.

## Further reading

- [Tell Google about localized versions of your page (Google Search Central)](https://developers.google.com/search/docs/specialty/international/localized-versions): The three ways to declare hreflang, the supported codes, x-default and the self and return link rules.

- [Managing multi-regional and multilingual sites (Google Search Central)](https://developers.google.com/search/docs/specialty/international/managing-multi-regional-sites): Country domains, subdomains, folders and parameters compared, and how Google picks a version.

- [Does duplicate content hurt SEO and AI search visibility? (Bing Webmaster Blog)](https://blogs.bing.com/webmaster/December-2025/Does-Duplicate-Content-Hurt-SEO-and-AI-Search-Visibility): Bing's December 2025 guidance on localized pages, hreflang and canonicals.

- [Sitemap in robots.txt: a third of Polish files omit it](https://www.outofplace.space/blog/sitemaps-robots-txt-polish-web): Our study of how Polish sites point crawlers to their sitemaps in robots.txt.
