International SEO Audits: Hreflang and Country Signals

International SEO audits check whether each market's page is the one search engines can find, trust and show to the right searcher. They compare URL ownership, hreflang pairs, canonicals, indexability and local content, then test what real buyers in each country see. Fix routing first, then localized content, then measure by market.

Most global sites share one brand and one template, while buyers in each country ask different questions in their own words. Prime AI Visibility saves answers for chosen questions, and this guide shows how to join that evidence to a technical audit with a URL-to-buyer matrix.

Who is this guide for?

It is written for a global marketing lead, an international SEO owner, a localization manager, a CTO and the country marketers who live with the results. International SEO audits suit a small team with one spreadsheet and an enterprise with dozens of locales.

Key takeaways:

• Language and country are different targets, and the audit has to say which one each URL serves.
• Hreflang only routes correctly when URLs, canonicals and local content agree.
• A translation is not a localized buyer answer. Test the questions each market asks.
• Measure every market separately, with the same method each time.

What do international SEO audits cover?

They cover five layers, in this order:

1. Scope. Which markets, languages and page types are in the audit.
2. Address. Which domain, folder or subdomain owns each market.
3. Routing. Hreflang pairs, x-default, canonicals and redirects.
4. Indexability. Whether each locale URL can be crawled and indexed.
5. Evidence. Local search results, local answers and local performance data.

The first four are technical and can be checked from your own site. The fifth needs observation in each market. An audit that stops at layer four can show every tag as valid while buyers in one country still see the wrong page.

What is the Global URL-to-Buyer Map?

This is the framework in this guide, six steps in order:

1. Markets. List each country and the business reason for it.
2. Query language. Record the language and wording buyers use there.
3. Address ownership. Assign one URL space to each market.
4. Hreflang and canonical reciprocity. Every alternate points back, and every URL canonicalizes to itself.
5. Localized evidence. Check that the page answers the local question with local facts.
6. Region-specific tracking. Report search and answer results per market.

Each step needs the previous one. A tracking report for a market whose pages are not indexable only measures the problem.

How do you scope the markets and pages?

Start from commercial markets, not from a list of every locale the site could support. For each market, write the country, the language or languages, the buying team's questions and the page types that matter, such as product, pricing, contact and top articles. A market with no owner in the country and no buyer questions written down is a candidate for removal from the audit.

Then build the URL matrix. One row per page, one column per market, and a cell holding the live URL, its canonical and its hreflang value. A generic example:

Page	en-us	en-gb	de-de	sv-se
Home	/us/	/gb/	/de/	/se/
Pricing	/us/pricing	/gb/pricing	/de/preise	/se/priser
Guide	/us/guide	/gb/guide	Not translated	/se/guide

The matrix is a teaching sample, not a real site. The empty cell is the point: a page with no German version must not appear in any German hreflang set, and the German visitor should land on a fallback you chose on purpose.

Language or country: which one are you targeting?

Google's documentation separates the two. A multilingual site offers content in more than one language. A multi-regional site explicitly targets users in different countries, and some sites are both. The difference decides the audit:

• A language target (de) serves German speakers wherever they are.
• A country target (de-de) serves German speakers in Germany.
• A language with regional differences (en-gb and en-us) needs different currency, spelling, units and legal wording, not just a flag.

Google also states that it does not use hreflang or the HTML lang attribute to detect the language of a page. It uses the visible content, so a page with a German tag and English text is read as English. Make the visible language obvious, and avoid a side-by-side translation on one page.

How do ccTLDs, subfolders and subdomains compare?

Google's documentation lists these options with their trade-offs, summarized here:

Structure	Example	Strength	Limit
Country domain	example.de	Clear geotargeting, easy separation	Cost, more infrastructure, can target one country only
Subdomain	de.example.com	Easy to set up, flexible hosting	Users may not read "de" as language or country
Subfolder	example.com/de/	Low maintenance on one host	Separation is harder, single server location
URL parameter	site.com?loc=de	Not recommended	Hard to segment, unclear to users

For a small team, a subfolder structure on one host is usually the lowest effort. For an enterprise with separate legal entities, country domains or subdomains can match how the business is run. International SEO audits do not pick a winner. It checks that the choice is applied the same way across every market, because a mixed structure is where most routing conflicts start.

Your URL architecture is part of the customer experience as well as a developer preference, so decide it with the country teams. Changing it later means redirects and fresh signals in every market.

What should hreflang and canonical tags look like?

Hreflang says which URLs are alternates of each other for different languages or regions. A canonical tag says which URL is the preferred version of one page. They do different jobs and must not contradict each other.

Google's guidance has four rules that matter most in an audit:

• Each language version lists itself and all other versions.
• Alternate URLs are fully qualified, with the scheme.
• If two pages do not point to each other, the tags are ignored.
• A generic fallback, marked x-default, helps unmatched users, such as a language selector page.

A generic example for the home page of the matrix above, placed in the head of every variant:

<link rel="alternate"
hreflang="en-gb"
href="https://example.com/gb/">
<link rel="alternate"
hreflang="en-us"
href="https://example.com/us/">
<link rel="alternate"
hreflang="de-de"
href="https://example.com/de/">
<link rel="alternate"
hreflang="sv-se"
href="https://example.com/se/">
<link rel="alternate"
hreflang="x-default"
href="https://example.com/">

On the UK home page the canonical points to the UK URL:

<link rel="canonical"
href="https://example.com/gb/">

The error to look for is a canonical on the UK page that points to the US page. That tells search engines to treat the UK page as a copy of the US one, and the hreflang pair then points at a URL that was told to defer. Every locale URL should be self-canonical.

What is x-default?

It is the hreflang value for users whose language or region you did not list. It usually points to a selector page or to a general version of the site. Google describes it as a catch-all for unmatched users. It is not a replacement for the listed locales, and it should itself be a live, indexable page that links to the other versions.

Does English need different content for the UK and the US?

Often yes. Google notes that localized versions are only treated as duplicates when the main content stays untranslated. English pages with small regional variations are a documented use for hreflang, such as content for the US, GB and Ireland. Differences that change a buyer's decision belong on the page: price and currency, spelling of product terms, units, delivery, regulation and the names of local partners.

If the only change is a currency symbol, hreflang still helps. If the buyers ask different questions, the page needs different content. One translation does not suit a finance director, an engineer and a country manager, so write the page for the question each of them asks in that market.

How do you diagnose mixed-language results?

When the wrong language shows up in a market, check causes in this order:

1. Indexing. Is the right locale URL indexed? Use Google Search Console URL inspection.
2. Reciprocity. Do the hreflang pairs return? Missing return links make annotations unreliable.
3. Canonicals. Does any locale canonical point to a different locale?
4. Visible language. Is the main content truly in the target language, with navigation to match?
5. Redirects. Does a script redirect users by IP or browser language? Google advises against automatic redirects between language versions, and its crawler usually has no language header and often comes from the United States.
6. Demand. Is the local page simply weaker than another page that answers the question?

Ahrefs documents its own hreflang error checks in its Site Audit help article, which is a useful list of what a crawler flags. Use Search Console's performance report filtered by country and query to see which URL gets impressions in which market. Average position in that report is an average across exposures, not the rank for one query at one time.

What do AI answers add to an international audit?

Search results are one surface. Buyers in each country also ask AI tools, and answers can differ by language, country and the sources available in that language. An audit can record what each tool displayed for a fixed question, in a stated market, and which source links it showed.

Limits matter here. A result depends on the tool, the settings and the day, and repeated checks do not guarantee what a real buyer sees. Google states that SEO fundamentals remain relevant to AI features and that no special markup is needed for them, so a clean locale structure is also the foundation for answers. Nothing here promises a placement in Google or ChatGPT.

Use a method like how to run an AI visibility audit for each market, with the same questions translated for meaning, not word for word. Record the country, language, tool and date next to every answer, and keep markets apart.

Prime AI Visibility lets you choose the questions and AI tools and saves each answer with its links. Its how it works page and its platform features page describe markets as part of the setup. A Comparable change only compares sufficiently aligned questions, tools, persona, market and measurement method. Read the measurement metrics page for the labels it uses.

How do you test regional AI answers?

1. Write 10 buyer questions per market in that market's language, from sales calls and local search data.
2. Keep one list per market. Do not mix languages in one list.
3. Run each question in each tool the same number of times per period.
4. Save the full answer, the source links and the date.
5. Mark whether your brand was named, whether your site was cited, and which locale URL was cited.
6. Compare each market with its own earlier results, not with another market.

A cited URL from the wrong locale is an audit finding: it points back to the matrix. Use AI search share of voice to define the share you report, and keep it per market. Whether you use a spreadsheet or Prime AI Visibility, the rule is the same: one saved record per answer, labelled with its market.

For the Spanish-language starting point, the herramientas AEO y GEO básicas guide lists free tools. For the strategic frame, read the AI visibility strategy guide, and for markup on locale pages, structured data for AI search.

What is the country-by-country gap matrix?

After the technical and evidence checks, summarize each market in one table with an owner:

Market	Routing	Content fit	Local evidence	Owner
en-us	Clean	Strong	Sampled	Country lead
en-gb	Canonical conflict	Partial	Not sampled	SEO lead
de-de	Missing return links	Not translated	Not sampled	Localization
sv-se	Clean	Partial	Sampled	Country lead

This is a generic sample, with word ratings instead of invented numbers. A gap matrix is useful only when each row has one owner and one next action.

How do you prioritize fixes from international SEO audits and test for regressions?

Rank fixes by effort and by the markets they unblock:

1. Routing errors first. A wrong canonical or a missing return link affects every page in the set.
2. Indexability next. Blocked or redirected locale URLs cannot rank at all.
3. Content gaps by revenue. Localize the pages that feed the main buying path.
4. Tracking last. Report by market once the pages are stable.

Effort is easier to size than results. A site with 4 markets and 25 templates has 100 locale pages to check, and an automated crawl can flag missing return links and cross-locale canonicals across all of them. Reading content fit takes a human per market. Treat international SEO audits as recurring work, with a crawl after each release, so a template change cannot silently break the pairs again.

Duplicate content is not a manual penalty by default. Google's documentation says localized versions are duplicates only when the main content stays untranslated, and the practical risk is that the wrong page is chosen, not a punishment.

What does a published international example show?

Percepture is a global digital marketing firm with a working relationship with Prime AI Visibility. Its page on technical SEO audits says it has done technical SEO and implementation work for global brands including Telstra and Alight.

For Telstra, that page covers the SG, HK, UK and USA sites and reports results after implementation: a 212 percent increase for the USA site against the period before Percepture, 14 percent for the UK site against the previous period, and average bounce rates of 6 to 8 percent across regions. Those are Percepture's own reported figures, and the two traffic numbers use different comparison periods, so they cannot be set side by side. The point for international SEO audits is the structure: four markets, one scope each, and a named comparator for every result.

For a client engagement, document international SEO audits by market scope, method and dated outcome. Keep the country-level URL lists and individual findings private, and publish only the scope and the result you have permission to share.

Which approach fits a small team and which an enterprise?

Team	Approach	Watch for
Small team	One subfolder host, a short matrix, a crawl each quarter	Skipping local question tests
Mid-size	A shared matrix, named owners per market, monthly checks	Different standards per country
Enterprise	Mixed domains with central rules, release-time regression tests	Template changes breaking pairs

An outside provider can run the audit or the implementation. Percepture offers generative engine optimization services, and its audit page describes technical work for international sites. To compare monitoring products for the answer side, use compare AI search trackers. If your buyers are other businesses, the AI search optimization for B2B service firms guide covers the buyer questions.

What should a skeptical buyer ask a consultant or tool vendor?

• Which markets and languages are in scope, and which are not?
• How do you check return links and cross-locale canonicals across the whole site?
• What does the report show per country, and what comparison period does each number use?
• Which of these findings were observed, and which were inferred?
• Who owns each fix after the audit?
• How do you test local AI answers, and with which settings?

What is the decision checklist?

• Each market has a business reason and an owner.
• The URL structure is the same across markets, or the exceptions are written down.
• Every locale URL is self-canonical and indexable.
• Every hreflang set includes the page itself and returns from each alternate.
• x-default points to a live fallback.
• The visible language matches the target on every page.
• Local questions are tested per market with saved answers.
• Reports are separate for each market.

What not to believe: that hreflang alone fixes wrong-country results, that one translation serves every buyer, that duplicate content means a penalty, that average position is a rank, and any promise of a fixed Google or ChatGPT placement.