Free hreflang tool
Check the hreflang tags on live pages for missing return links, missing self references, invalid codes and canonical conflicts, with the exact tag to add for each problem. Or generate a correct set for a new page.
The run hit its limit of 25 URLs or 45 seconds, so some alternates were not fetched. Results for those are marked as not checked.
Errors stop Google trusting the set. Warnings may be deliberate and need a decision. Each fix is the exact tag to add to the page named.
Open a row to see the tags found on that page and the result of each check.
| URL | hreflang | Status | Self | Return | Canonical |
|---|
Build one complete tag block from the pages found, with every page listed once, ready for the head of each page.
Paste the whole block into the <head> of every page in the set. Each page lists itself and all the others. A page that lists only the other versions has its tags ignored.
Each check follows Google's documentation for localised pages. There is no score: a set of hreflang tags is either valid or it is not. When a page fails a check, the result names the page and shows the tag to add.
Google will not trust the set until these are fixed.
Often deliberate. Check that the setup is what you intended.
Reported for information. Not a fault.
The checker reads the raw HTML head and the response headers of each URL. It does not run JavaScript, so tags added by scripts are not seen, and it does not read XML sitemaps yet. Only the URLs you enter and the alternates they declare are fetched. Nothing is stored.
These are the labels used by audit tools such as Semrush and Ahrefs. In most cases the fix is one line in one template.
en-UK, ja-JA, zh-CH and en-EU are all invalid. The language must be ISO 639-1 and the region ISO 3166-1: en-GB, ja-JP, zh-CN. Google states that EU, UN and UK have no effect.hreflang="HK" is invalid. A region can only follow a language: zh-HK, en-HK. Add one entry per language you publish for that region.div can close the head early, and tags injected by scripts after load are unreliable./hk/, not the http version, and not a URL that redirects to a trailing slash.A value has up to three parts, always in this order. Only the language is required. The script and region narrow down who should see the page.
zhRequiredISO 639-1. Two letters, lowercase. Google specifies this list, so three-letter codes and withdrawn codes such as iw are flagged.
HantOptionalISO 15924. Four letters. Used mainly for Chinese (Hant, Hans) and for languages written in more than one alphabet.
HKOptionalISO 3166-1 alpha-2. Two letters, uppercase. It cannot be used on its own, and values such as UK, EU and UN have no effect in Google Search.
Combined: zh-Hant-HK. Language tags are not case-sensitive, so zh-hant-hk also works, but the conventional spelling is easier to read and to search for in templates.
Choosing a code
en, ja, zh-Hant.en-HK, en-SG, ja-JP.en-SG, zh-SG, ms-SG and ta-SG, each pointing at its own translation.en-HK and en-SG but no plain en, readers elsewhere get whichever page Google prefers. Add an x-default so that choice is yours.lang attribute.Examples
en-UK is not a valid codeja-JA is not a valid codezh-CH is Switzerlanden-EU is not accepted because the EU is not a countryCanonical says which URL to index when several are the same page. Hreflang says which indexed URL to show to which reader. Most multilingual sites need both, and they have to agree.
A Japanese page is not a duplicate of its English original. Each version keeps a self-referencing canonical, and the hreflang set links them. If the Japanese page canonicalises to the English one, it drops out of Japanese results.
An en-HK page and an en-SG page with the same copy are duplicates to Google. Its guidance is to pick a preferred version with rel=canonical and keep the hreflang tags, so the right regional URL can still be served. The canonical should stay in the same language.
Tracking parameters, capitalised paths and http versions should canonicalise to the clean URL, and the hreflang set should only reference that clean URL. If the checker shows a canonical that differs from the hreflang URL, fix that first.
Google supports HTML tags, HTTP headers and sitemap entries, and says there is no benefit in Search from using more than one, only more to manage. Use one.
The most common method, and what the generator on this page produces. The same block goes in the head of every page in the set. Easy to inspect, but each template has to be kept in sync.
Suited to large sites and to teams that cannot edit templates. Each URL entry lists every alternate, including itself, using the xhtml namespace. One generated file covers the whole site. The checker does not read sitemaps yet.
For files without an HTML head, such as PDFs. The server sends the set in a Link header. The checker reads these headers and reports them alongside HTML tags.
It tells search engines that a group of URLs are versions of one page for different languages or regions, so they can show each reader the right one. Google recommends declaring the relationship wherever localised versions exist. It translates nothing, does not decide the page's language, and does not lift rankings by itself. It affects which version is shown.
Usually. Canonical handles duplicates of one page; hreflang handles alternates of one page for different readers. Translated pages each carry a self canonical plus the shared hreflang set. Same-language regional copies are the special case: decide whether they should be indexed separately, then set the canonical to match rather than leaving it to the CMS default.
Yes. en is English for anyone. en-HK adds a Hong Kong signal, which matters when the page carries Hong Kong pricing, delivery or legal content. If the page is the same for everyone, use en and leave the region out.
Choose by reader, not by writing system. Both are usually Traditional Chinese, but Hong Kong and Taiwan differ in vocabulary, currency and what people search for. A page written for Hong Kong is zh-HK; for Taiwan, zh-TW. One Traditional Chinese page serving both is zh-Hant. Google detects the actual language from the content, so no tag can make a Simplified page count as Traditional.
No. It names the page for readers who match none of the listed versions, and Google says it works best for language selector pages. Useful when you have a global home page or a chooser; unnecessary when every reader matches one of your versions. More on x-default above.
Yes. Google accepts HTML link tags, HTTP Link headers and sitemap xhtml:link entries, and says using more than one brings no benefit in Search. Pick one method for the whole site. This checker reads the first two today; sitemap checking is planned.
Usually the return tag names a slightly different URL: http rather than https, a missing trailing slash, www against a bare domain. Or the tag sits outside the head, or is added by JavaScript after load. Run the pair through the checker; it compares the exact final URLs and shows the tag it found on each page.
International SEO
Hreflang is one part of a wider international setup. Site architecture, localisation, canonical decisions and market-specific search demand need to work together. That is the work I do across Hong Kong, Singapore, Japan and beyond.
Explore International SEOOr ask about your site