Global SEO Localisation

1. Missing return tag. Hreflang must be reciprocal. If /en/ points to /fr/ with hreflang="fr", then /fr/ must point back to /en/ with hreflang="en". A single missing return tag invalidates the cluster — Google treats both pages as if no hreflang were set.
2. Wrong region codes. Spanish for Spain is es-ES. Spanish for Mexico is es-MX. Many sites ship es-SP or spanish; both are invalid and silently ignored.
3. Self-referential mismatch. Every hreflang cluster must include a self-referential tag (the French page must include hreflang="fr" pointing at itself). Skip it and the cluster is broken.
4. Mixing protocols. A site that serves https://example.com/fr/ but lists http://example.com/fr/ in hreflang invalidates the tag.
5. Conflicts with canonical. If a page canonicalises to a different language version, hreflang is overridden. Common on sites that canonicalise everything to the English root.
6. x-default missing. The hreflang="x-default" tag is what Google falls back to when no language matches. Most sites omit it; the result is poor matching for users on uncommon locales.

We crawl every URL, build the hreflang graph, validate reciprocity, check status codes (a hreflang pointing at a 301 chain is also a bug), and verify implementation matches the sitemap. The output is a defect list with the exact line of HTML or XML to change. We fix the issues before any new content goes live — fixing translations on top of broken hreflang is paying twice for the same problem.

For most B2B and DTC clients, we recommend subfolders. The backlink inheritance is the deciding factor — building international link equity from scratch costs more than the SEO penalty of not using a ccTLD. We make exceptions for brands targeting Germany (where .de trust is meaningfully higher) and any brand targeting China, where a .cn domain and ICP filing are practical necessities.

Migrating from subdomain to subfolder (or vice versa) is one of the highest-risk SEO operations there is. If you are already live in one structure and considering a change, we will walk through the cost-benefit before quoting; sometimes the answer is to stay put and concentrate on content depth.

• A direct answer in the first 100 words. AI engines extract the gist of the page from the opening. If your French homepage opens with brand poetry — "Welcome to Acme, the world's leader in..." — you will not be cited. If it opens with "Acme is a Paris-based supplier of X, serving Y industries, with Z certifications", you will.
• Clear factual statements. AI engines distrust vague claims. "Trusted by thousands" is invisible; "Trusted by 312 UK law firms since 2014" is citable.
• Structured data. Service schema, FAQPage schema, BreadcrumbList — all give AI engines reliable hooks for extraction.
• Original data. Pages that contain a statistic, benchmark, or table the AI hasn't seen elsewhere are disproportionately cited.
• Per-locale structure. A German user gets cited a German source. A page in French with broken French is invisible.

Every page we translate is re-architected so the first 100 words contain a direct, factual answer to the page's primary search intent. In the target language. With the brand name, location, certifications, and a single concrete differentiator. We add Service and FAQPage schema; we cite our own sources where applicable; we publish an "Updated [date]" stamp on every long-form page (this is one of Google's freshness signals and one of the few honest ranking factors left).

We will not promise "AI ranking guarantees" because they do not exist. AI Overview citation is a side effect of being the best-engineered page for the query in the target language — not a quirk you can game. We rebuild pages to deserve citation; what happens after that depends on the query landscape.

We integrate with WPML (the dominant choice), Polylang, and Weglot. WPML's String Translation module is our default for theme strings and plugin output; the Translation Management module handles posts and pages. We push translations via WPML's API so authors do not see translation pull requests in their post editor.

Shopify Markets is the right choice for most stores launching internationally. We translate product titles, descriptions, metafields, checkout strings, email templates, and theme strings, then sync via the Shopify Translation API. RTL stores need theme work in addition to translation — we deliver both, with a tested Arabic checkout flow.

Webflow's native localisation released in 2024 is now mature enough to use; before that we worked through Weglot. We push via the Webflow CMS API for collection items and through Designer for static strings. Webflow's hreflang implementation requires manual verification on every site — we do that as part of the launch pack.

Headless platforms are the cleanest integration path. Each has a locales API that lets us push translations as a per-locale field on every entry. We work with your developers to make sure the front-end consumes the right locale, hreflang renders correctly, and SSR caches per locale.

If your CMS is custom-built, we work to your API. Most projects need a simple POST /api/translations endpoint that accepts a payload of {locale, key, value} tuples. We can also work file-based (CSV or JSON drops to a Git repo or S3 bucket). We adapt; you do not refactor.

• Product catalogue — titles, descriptions, attributes, metafields. For large catalogues (1,000+ SKUs) we use a CSV pipeline and a glossary so terminology is consistent across the entire range.
• Checkout — payment method labels, shipping copy, returns policy, T&C. Local payment methods (iDEAL in NL, Klarna in DE, Mada in SA, Pix in BR) need to be visible above the fold or conversion drops.
• Currency and tax — we work with your finance team on local currency display, VAT/tax-inclusive pricing, and the small typographical conventions (German uses 1.234,56 €, French uses 1 234,56 €).
• Trust signals — Trustpilot widgets per locale, local review aggregators (Trusted Shops in DE, Avis Vérifiés in FR), guarantee copy.
• Email sequences — order confirmation, shipping notification, abandoned cart, post-purchase. The abandoned-cart email is the highest-ROI translation in e-commerce and routinely ships in English to French buyers.
• RTL storefronts — for Arabic and Persian markets, the entire layout needs mirroring. We have shipped this for several DTC brands targeting GCC.

For e-commerce we report two numbers per locale: organic traffic growth and conversion rate vs the English baseline. If conversion is meaningfully lower than English, that is a localisation problem (or a payment-method problem, or a trust-signal problem) — and we fix it. If conversion matches English, the localisation has done its job and the lever shifts to traffic acquisition.

• Organic sessions per locale from GA4 — month-over-month and year-over-year.
• Keyword positions per locale from Google Search Console — top 10 changes, biggest gainers, biggest losers, and the keywords your competitors gained that you did not.
• Pages with zero traffic — if a page we translated has zero organic sessions after 90 days, the localisation failed and we fix it under the retainer.
• Conversion rate per locale — for e-commerce and SaaS sign-up sites, this is the headline number. We watch for divergence from the English baseline as the early-warning signal.
• AI Overview citation rate — sampled monthly across your top 50 keywords per market. This is the new top-of-funnel metric and currently uncovered by most agencies' reports.

• Word count delivered.
• "Translation memory leverage savings" — these are real but they don't grow your funnel.
• Vanity metrics like impressions without click-through context.

You get the report monthly. If a target number isn't moving, we propose a remediation in the same email — usually a content depth pass, a hreflang fix, or a refresh of an underperforming page. Nothing waits 90 days for a quarterly review.