# International SEO for SaaS: localisation Hreflang, ccTLD vs subfolder, localising keywords and pricing, and what ChatGPT says in German, French and Spanish, measured. We automate it. Canonical: https://ignura.com/localization Last updated: 2026-10-27 Author: Ignura ## In short - Google needs to know which page is for which language. Hreflang tells it: each version must list itself and every other version, and Google ignores tags that don’t point both ways. - Google works out a page’s language from its visible text, not from code. A translated page has to be translated top to bottom. - AI answers in other languages read English pages. Asked the same buyer question in German, French and Spanish, ChatGPT named the same five tools, ran its searches in English and cited only English third-party pages. - That’s an opening: in many categories nobody has written the local comparison yet. Start with a free Growth Check. ## Subfolder, subdomain or country domain Google lists the options: a country domain (example.de), a subdomain (de.example.com), a subfolder (example.com/de/), or URL parameters, which it doesn’t recommend. Country domains are clear but cost more and target one country each. Subfolders are easy to set up and run on one host. For most B2B SaaS we use subfolders: one domain, one set of links, one place to maintain. ## Hreflang without the usual bugs Each language version lists itself and every other version, the links go both ways, and an x-default catches everyone else. Google says it doesn’t use hreflang or the lang attribute to detect a page’s language; it reads the page. The bugs we find most: missing return links, hreflang pointing at redirected or noindexed pages, and a language switch that only works with JavaScript. ## Localise, don’t translate The words people search change by market, so we research them in each language instead of translating the English list. Examples, customers, regulations and prices change too. Show the currency you charge in, and the rules buyers there care about. ## What the AI engines do abroad In our test ChatGPT answered in German, French and Spanish, but every web search it ran was in English and every third-party page it cited was in English. Only 5 of the 13 pages it cited in Germany were also cited in the US answer. So two things matter: being on the English pages the engines read everywhere, and being first with a good page in the local language. One test can’t tell you how your category behaves, so we ask your questions in your markets. **0** third-party pages in German, French or Spanish were cited for the question in those markets. ## Which markets first We start from your own numbers: visits and sign-ups by country in analytics, impressions by country in Search Console. Then where rivals have no local pages, where you can sell and support, and where buyers search in their own language. We run this site in six languages ourselves: English, Spanish, French, German, Brazilian Portuguese and Japanese, with hreflang and one sitemap per language. ## Mistakes we see - **Raw machine translation.** Readers notice in a sentence, and so does any buyer comparing tools. - **Hreflang without return links.** Google ignores tags that don’t point both ways. - **Translating the English keyword list.** People abroad search in their own words, not yours. - **Versions behind a pop-up or an IP redirect.** If people and crawlers can’t reach a version by a link, it may as well not exist. ## What our AI does - Asks your buyer questions in each market’s language, from that market, on ChatGPT search and Perplexity. - Checks every hreflang pair on your site for return links and live targets. - Pulls search demand and rankings per market. - Flags pages translated in the menu but still English in the body. ## What we do - Choose markets with you, from your own numbers. - Research local search terms with native speakers. - Write or edit local pages with native speakers. No raw machine output goes live. - Set up the URL structure and hreflang with your engineers. ## The first four weeks 1. Your markets ranked from your own data, and your buyer questions asked in each language. 2. URL structure and hreflang set up or fixed. 3. The first local pages: pricing, who it’s for, the strongest comparison. 4. Local rankings and AI answers measured again, and the next market planned. ## The deal Named in 5 of your 20 buyer questions by week 4, or you don’t pay the second half. $8,000 for four weeks, half to start. Then $4,000 a month, with a month’s notice to stop. ## Questions ### Country domain or subfolder? Subfolder for most B2B SaaS: one domain, one link profile, easy to run. Country domains make sense when you’re a different business in each country. ### Is translated content a problem for Google? Google cares whether a page is useful. Its scaled content policy covers pages made mainly to rank, however they’re produced. An accurate, edited translation for real buyers is fine. ### Do AI engines name different tools by language? In our SOC 2 test, no: the same five in all four markets, though it read different pages in each. Your category may differ, so we test it. ### Which languages can you do? We run Ignura in six ourselves. Tell us your markets on the call and we’ll say which we can do well. ## Sources - Ignura: [Six SOC 2 buyer questions on ChatGPT search, Perplexity and Google, 27 Oct 2026 (the data)](https://ignura.com/data/search-2026-10-27.json) - Google Search Central: [Tell Google about localized versions of your page](https://developers.google.com/search/docs/specialty/international/localized-versions) - Google Search Central: [Managing multi-regional and multilingual sites](https://developers.google.com/search/docs/specialty/international/managing-multi-regional-sites) - Google Search Central: [March 2024 core update and new spam policies](https://developers.google.com/search/blog/2024/03/core-update-spam-policies) Book a 15-minute call about your website: https://ignura.com/book