9 July 2026
Framer Multilingual SEO: Getting Indexed in Every Language
Translating a Framer site is the easy half. What Google actually requires, what it explicitly ignores, and what its spam policy really says about machine translation since March 2024.
Most multilingual Framer sites fail at the same place. The translation is fine. The pages exist. And Google either ignores them, or serves the German page to a French visitor.
Translation is the easy half. What follows is the other half, with what Google actually says rather than what SEO articles repeat.
First, three things Google explicitly does not do
Getting these wrong wastes weeks.
Google does not read your lang attribute. Its documentation is unambiguous: “Google doesn’t use hreflang or the HTML lang attribute to detect the language of a page; instead, we use algorithms to determine the language.” Setting lang="fr" is worth doing, but for accessibility (screen readers need it, and WCAG 3.1.1 requires it at Level A) and for browser translation prompts. It is not an SEO signal. Anyone telling you otherwise is repeating a myth.
Google does not rank you higher for having hreflang. hreflang swaps the URL that already ranks for the locale-appropriate one. It does not promote anything.
Google does not treat translated pages as duplicate content. Verbatim: “Localized versions of a page are only considered duplicates if the main content of the page remains untranslated.” Translation is transformative. The only duplicate risk is translating your navigation and leaving the body in the original language.
URL structure, and the preference Google never stated
Google documents four structures and, contrary to received wisdom, states no preference among the first three.
- ccTLD (
example.de): clear geotargeting, but expensive, more infrastructure, and it can only target a single country. - Subdomain (
de.example.com): easy to set up, allows different server locations, but users may not recognise the geotargeting from the URL. - Subdirectory (
example.com/de/): easy to set up, low maintenance, single server location, harder to separate sites. - URL parameters (
example.com?loc=de): the only one Google explicitly labels “not recommended.”
The practitioner consensus that subdirectories are best, because they inherit the root domain’s authority, is real consensus, and it is not a Google statement. Say so, and choose on your own constraints. A Framer site with translated pages at /fr and /de is on the subdirectory path, which is the low-maintenance choice.
One thing that is a Google statement: geotargeting is about country, not language. Language targeting is hreflang’s job.
What third-party translation services change
Services that serve your translated pages from their own infrastructure can be indexed correctly, and Google has never endorsed or condemned any of them by name. Do not believe any vendor claiming Google approves of their approach; there is no such document.
What matters is mechanical, and it follows from Google’s general guidance:
- The translated pages need their own crawlable URLs and server-rendered HTML. A JavaScript layer that swaps content on a single URL will not get its translations indexed as separate pages.
- If a service showed translated content to users and the original to Googlebot based on user-agent, that would be cloaking, which Google’s spam policy defines as “presenting different content to users and search engines with the intent to manipulate search rankings.” Reputable services avoid this.
Weglot serves translated versions on a subdomain or a subdirectory, with a reverse proxy available on Enterprise. Linguana serves translated static pages from its own infrastructure on subdirectories. Both are indexable. The question is whose URLs accumulate the authority, and what survives if you stop paying.
What machine translation actually does to your rankings, in 2026
This is where almost every article is out of date, including an earlier version of this one.
Until early 2024, Google’s spam policies contained a section called “Spammy automatically-generated content” that listed, as an explicit example: “Text translated by an automated tool without human review or curation before publishing.”
On 5 March 2024, Google replaced that section with a new policy, “scaled content abuse.” Its own announcement says the new policy “builds on our previous spam policy about automatically-generated content, ensuring that we can take action on scaled content abuse as needed, no matter whether content is produced through automation, human efforts, or some combination.”
The standalone bullet about unreviewed translation is gone. In the current spam policy, translation appears only once, as an example of scaled abuse: “scraping feeds, search results, or other content to generate many pages (including through automated transformations like synonymizing, translating, or other obfuscation techniques), where little value is provided to users.”
So the honest 2026 position:
Machine translation is not spam. It becomes spam when it is part of generating many pages whose primary purpose is manipulating rankings rather than helping users, however those pages were made.
The test moved from “was a human involved?” to “was this made at scale to manipulate rankings, and does it add value?” Translating your fifteen real pages with AI, and reading them, is nowhere near that line.
That said, review still pays, and there is evidence for where. ISO 18587, the standard governing post-editing of machine translation, names the recurring failure classes: literal translations, grammar mistakes such as negations, and names that should not have been translated. Peer-reviewed work (EAMT 2020) documents machine output that was accurate and grammatical yet failed a brand’s formality requirements outright, and product names translated into nonsense. Gender defaults are a well-studied failure too.
The practical rule: translate with AI, then read the pages that convert. Your homepage, your pricing page, the page your ads point at. Put brand names and jargon in a glossary before you start, because neither DeepL nor Google Translate has a do-not-translate concept by default; both implement it as a glossary entry mapping the term to itself.
The five things that get you indexed
1. Each translated page is a real, crawlable URL returning HTML with the translated content in it. Nothing else on this list matters otherwise.
2. hreflang on every language version, reciprocal, self-referencing, absolute. This is where sites break, and Search Console stopped reporting the errors in 2022. The dedicated hreflang guide covers the rules and how to verify.
3. Each language version canonicalises to itself. A French page whose canonical points at the English one has asked not to be indexed.
4. A language switcher made of real links. Google is explicit: “Google can only crawl your link if it’s an <a> HTML element with an href attribute.” It says it cannot reliably follow <span href>, <a onclick>, or href="javascript:goTo()". A switcher built from buttons and JavaScript can leave your other languages undiscovered.
5. No automatic redirection by IP or Accept-Language. Google: “Avoid automatically redirecting users based on their perceived language. These redirections could prevent users (and search engines) from viewing all the versions of your site.” Its recommended alternative is exactly point 4: “consider adding hyperlinks to other language versions of a page.” If you want to help, show a banner suggesting the other language. Do not force it.
Two more that nobody automates for you: give each translated page its own title and meta description in its own language (Google publishes no rule here, but a French page carrying an English snippet is a poor match for French searchers), and make sure every language version appears in your sitemap. Framer generates the sitemap for you and includes locale versions when you use native Locales.
Don’t forget the CMS
If your blog or product catalogue lives in a CMS collection, translating the static pages leaves most of your indexable content in one language. Collections have their own structure and their own URLs, which is what CMS Translator exists for. There is a guide for that.
Where Page Translator fits, honestly
Page Translator gives you point 1 (real pages at real URLs, in your own project), point 3, and a switcher made of real links.
On point 2, read carefully. The plugin injects hreflang tags with a script that runs in the browser, not into the served HTML, because Framer locks server-rendered lang and hreflang to its paid native Locales. Google renders JavaScript and can pick them up; Bing is less reliable at it. It translates page slugs only in DeepL mode, and never translates page titles or meta descriptions, which Framer’s native Locales and Linguana both handle.
If server-rendered hreflang, translated slugs and per-locale metadata are requirements, native Locales is the stronger tool, and it costs $20 per locale per month. That is a real trade, and you should make it with the facts.
Primary sources: Google Search Central, Managing multi-regional and multilingual sites (updated 10 December 2025); Tell Google about localized versions of your page (updated 22 December 2025); Spam policies (updated 15 May 2026); March 2024 core update and new spam policies; Make your links crawlable. ISO 18587:2017. Nunziatini & Marg, EAMT 2020. Checked 9 July 2026.