Colquhoun.AI
← All guides

Research Guides

Audit Hreflang as a Page Family, Not a Tag Count

· Colquhoun.shop editorial

A translated page can have perfectly formed language tags and still point visitors toward the wrong content. The useful unit of review is not one tag. It is a family of equivalent pages: the same guide, product or service, presented for different language or regional audiences.

This worksheet is for a small multilingual website whose owner can inspect its public pages. It separates three questions: does each version exist, do the annotations connect the right versions, and can people actually choose the version they need? It is a manual research method, not a claim that Colquhoun AI provides a working automated hreflang checker.

First decide whether the pages belong together

Hreflang describes localized alternatives. Start with one content item and list its real variants. A French product detail page is not equivalent to an English category page simply because both mention the same product.

Record the intended audience in words before writing a code. “English for readers in the United Kingdom” makes a useful requirement. “UK page” leaves the language unstated. Google's localized-versions documentation explains supported language and optional region codes, including en-GB for that example. Do not use a country code alone or invent a tag from a marketing region name.

If a translation does not exist, record a gap. Do not manufacture a regional URL or point the missing translation to an unrelated homepage to make a spreadsheet look complete. A smaller accurate family is more useful than a large misleading one.

Build one evidence row per version

Use these fields for each member of the family:

Do not place private preview URLs, session tokens or customer information in a worksheet intended for sharing. A public example or redacted finding is enough to explain the defect.

Compare sets rather than counting tags

Google supports HTML, HTTP-header and sitemap implementations. Its documentation treats these methods as equivalent and says there is no Search benefit to implementing all three. Use the method that fits the existing system; do not add a second mechanism merely to increase a checklist score.

For the chosen method, compare the complete set of alternatives across the family. Each language version should list itself and the other versions. Use fully qualified URLs, including the scheme, rather than relative paths. Check reciprocal references: when page A identifies B as its alternative, B should identify A.

A raw count misses important errors. Three tags might contain the same destination three times. Two pages might have equal counts but disagree about the German URL. The set comparison asks whether the intended relationships are actually present.

In an HTML implementation, inspect the head elements rather than assuming the visible language menu creates search annotations. In a header implementation, inspect the actual response headers. In a sitemap implementation, inspect the relevant entry and its language alternatives. Save the evidence from the mechanism being reviewed.

A fictional three-page finding

Imagine a delivery guide with English, French and German versions. English and French list all three URLs, but German lists only itself and English. All three pages return 200 and all contain the correct guide.

The first finding is not “German SEO failed.” It is: “The German entry omits the French alternative, so the French-to-German relationship lacks its return reference.” That wording states the observed defect without predicting rankings.

The proposed repair is to correct the German annotation set, then recheck all three members. If each of three versions lists all three, the family has nine declarations. Nine is a consistency check for this fictional family, not a universal SEO target. It says nothing about translation accuracy or whether Google will display a particular URL.

Now imagine that the French language-menu link leads to the site's French homepage instead of the delivery guide. That is a separate navigation defect. Fixing metadata would not fix the reader's journey. Give it its own action and verification result.

Keep language declarations and fallback choices distinct

The HTML lang attribute declares the language of a document's content. W3C's language-declaration guidance recommends setting it on the html element and using language attributes where text changes language. It is not a replacement for annotations connecting alternate URLs.

Google says it determines page language algorithmically rather than using hreflang or HTML lang to detect it. That does not make accurate HTML language markup unnecessary; the two mechanisms serve different purposes.

For visitors whose language is not explicitly covered, review whether an x-default alternative is appropriate. Google's localized-versions guidance describes it as a fallback for unmatched languages, particularly on language or country selectors. It is not an additional language and should not be used to disguise a missing translation.

Test the reader's choice before approving the change

Google's multilingual-site guidance advises avoiding automatic language redirects that can prevent people and search engines from reaching all versions. It also recommends links that let users choose another language.

Open each public version directly. Follow its language selector to another member and back again. Confirm that the content stays equivalent and that a language choice is not immediately undone. Do not change geolocation, security settings or account preferences merely to complete an SEO audit; record any access limit instead.

After an authorized release, repeat the worksheet against production, not just a preview. Confirm final URLs, annotation sets, return references and selector destinations. Keep the before-and-after evidence with the release record.

The final deliverable should be a small list of precise repairs, not a promise of more international traffic. Correct relationships make the website easier to interpret. Search visibility still requires separate observation, and business impact requires real measurement.

This guide describes the built-in Web Reader and Keyword Density behavior verified for this publication. Available data depends on the page response; these tools do not measure visits or search rankings.