SEO
AI translation for your website: does Google penalise machine-translated pages?
Copy for AI
No, Google doesn’t penalise a page just because a machine translated it.
So AI translation on your website isn’t a problem in itself. What Google does penalise is translated content published at scale without adding anything for the reader.
The verdict: the risk isn’t in the tool, it’s in what you do with it. A proofread translation with the right keywords can rank fine; an unreviewed copy of hundreds of pages is a risk.
Below: what Google’s rules literally say, what makes an SEO translation different, and how to set up a multilingual site that gets found in every language.
Related guides
Want the bigger picture first (machine, translator or hybrid)? Read our guide to website translation.
If you use a translation platform, our Weglot review is a useful addition.
Moving away from one? Read how to leave a translation tool without losing SEO.
What does Google say about AI translation and automatically translated content?
Google covers automatic translation in its spam policies under scaled content abuse: creating many pages mainly to manipulate rankings rather than help users.
In Google Search’s spam policies, the examples literally include scraping content to create many pages, “including through automated transformations like synonymizing, translating”, where this provides little value to users.
The qualification is at the end
Read that sentence carefully. The key part comes last: where little value is provided.
Google says the policy applies no matter how the content is created. So a translation in itself isn’t a violation.
For example, a site that rolls out its product catalogue in twelve languages without anyone checking the pages does enter the danger zone.
Google dropped its robots.txt advice
A second signal comes from the documentation itself.
In June 2025 Google removed from its guidelines for multilingual sites the old advice to block all automatically translated pages via robots.txt.
The reason, in Google’s overview of documentation changes: aligning with the March 2024 spam policy on scaled content.
In other words, the message used to be “hide machine translation”. Now it’s “judge every page on its value”.
Metadata counts too
You find the same line in Google’s guidance on AI-generated content: focus on accuracy, quality and relevance, especially with automatically created content.
That also applies to metadata:
- the title
- the meta description
- structured data
- alt texts
Those four are the ones most often forgotten in translation projects.
Same rules for AI-written text: see AI content and SEO.
What is the difference between a translation and an SEO translation?
An ordinary translation converts your text into another language. An SEO translation also makes sure the page gets found in that language.
The biggest difference lies in the keywords. People search with different words in every language, and they’re rarely the literal translation of your Dutch term.
An example from our own research
“Website vertalen” is searched around 170 times a month in Belgium.
The literal French counterpart “traduction site web” gets 40 searches among French-speaking Belgians, and “traduire site web” just as many. In English, “website translation” gets around 90 a month in Belgium (Google Ads data, October 2026).
Here are those four volumes side by side:
Anyone who only translates picks one by chance. Anyone who makes an SEO translation first checks which variant people search for and adapts the page.
Where to find the local keyword
Google’s Keyword Planner lets you set the location and language per plan. This is what a plan looks like, here for “seo tools” in the US:
Keyword Planner for “seo tools”, United States, plan from 17 October 2026.
Run the same plan per language and market before you translate a word. Our guide to local keyword research explains the method.
What changes in an SEO translation
What you take into account, element by element:
| Element | Ordinary translation | SEO translation |
|---|---|---|
| Keyword | Translated literally | Researched again per language and market |
| Title and meta description | Often forgotten or too long | Rewritten around the local keyword |
| URL (slug) | Stays in the source language | In the language of the page |
| Headings | Translated | Tuned to the questions asked locally |
| Alt texts and file names | Usually not | Translated |
| Internal links | Sometimes point to the source language | Point to the pages in the same language |
| Hreflang and canonical | Not applicable | Correct per language version |
| Proofreading | Optional | By a native speaker |
Our experience: the typical mistake is a tool picking the wrong keyword. For a time-tracking software client, Weglot translated “time registration” as “tijdsregistratie”, while Belgian searchers type “tijdregistratie”.
One letter, and the page targets a word its market doesn’t use.
How does Google recognise the language of your page?
Google looks at the visible text on the page, not at the code.
The guidelines for multilingual sites state that Google doesn’t use code-level information, such as the lang attribute or the URL, to determine the language.
Two practical consequences
That leads to two rules:
- One language per page. Keep content and navigation in one language; avoid side-by-side translations (NL and FR in two columns).
- Don’t translate only the boilerplate. Google calls a page with only the menu and footer translated a poor user experience.
You often see that second one with half-finished translation plugins.
The lang attribute still matters
That doesn’t make the lang attribute useless. Screen readers use it to pick the right pronunciation, so for accessibility it should be set correctly.
Each language counts on its own
The same holds for AI assistants. Ask ChatGPT a question in Dutch and it builds a separate Dutch answer:
ChatGPT free (temporary chat), 17 October 2026: three bullets, about 70 words, no sources.
Our experience: language models weigh each language separately.
So a multilingual rollout is a low-effort, high-impact lever.
How do you set up a translated site correctly from a technical point of view?
The technical setup decides whether Google finds your translated pages at all.
According to Google, Googlebot usually crawls from the United States and doesn’t send a language preference.
So a site that switches language by cookie or browser setting may show Googlebot only one version.
Four basic rules
That’s why four basic rules apply:
- A separate URL per language, for example
/fr/and/nl/or a separate domain. No language switching on the same URL. - Hreflang on every language version, reciprocal and self-referencing.
- No automatic redirects based on language. Let a language switcher link to the equivalent page, not the homepage.
- A self-referencing canonical per language version. A French page with a canonical to the Dutch version won’t be indexed.
Hreflang makes Google show the right version to the right searcher. The details are in hreflang tags explained, the classic mistakes in fixing hreflang errors.
Check after launch
After launch, check that every version is reachable and the tags point to each other correctly.
Our free hreflang checker shows which language versions exist, whether the hreflang is right and whether an automatic translation plugin is running.
Changed URLs along the way? Run a sample through the redirect checker too.
When is machine translation responsible?
Machine translation is a production tool, not an end product.
For SEO, it’s responsible as soon as a human reviews the pages that matter.
A practical split
Split your pages into three groups:
- Machine is enough (with spot checks): large catalogues, technical docs, low-pressure knowledge base articles.
- Machine plus proofreading: blog articles and informational pages, checked by a native speaker.
- Rewritten by a human: homepage, service pages, landing pages, CTAs and forms.
The native speaker in the middle group checks terminology, tone and the keyword.
The last group is where the visitor decides whether to get in touch. It’s also where a machine sentence gives you away fastest.
Our experience: we do all translations in-house. Our workflow is AI translation plus local search data for the right keyword, then a human pass where it’s really needed.
Why this hybrid works
That hybrid approach fits what Google asks: every published page must be useful and accurate.
It also fits what your customer asks. A French-speaking buyer forgives a missing translation sooner than one that sounds strange.
That’s the subject of our article on translating content for multiple markets. Our guide to AI vs human content weighs the same trade-off for writing.
Frequently asked questions about AI translation and SEO
Will I get a penalty for using Google Translate on my site?
Not automatically. Google assesses whether content was created at scale without value for the user, and names automatic translation as one of the techniques.
A few machine-translated pages that you review and improve aren’t a problem. Thousands of unreviewed translated pages are.
Should I set automatically translated pages to noindex?
Only if they’re not good enough to show a reader. Google recommends keeping content that falls under scaled content abuse out of the search results.
Better: improve the most important pages, and only exclude the weak remainder temporarily until it’s proofread.
Should I translate my URLs too?
Yes, if your platform allows it. A slug in the page’s language is clearer for the visitor and contains the local keyword.
If you change existing URLs, set up 301 redirects from the old to the new addresses, as described in our 301 redirect plan.
Is DeepL better than Google Translate for SEO?
For SEO, the engine matters less than the proofreading. Both deliver a usable first draft; neither knows your keywords, your sector or the Belgian context.
The difference between a page that ranks and one that doesn’t lies in what happens after the machine.
Can I export my translations from a translation plugin?
Usually yes, and it’s the first step to owning them.
See how to export translations from Weglot.
Want your translated pages to be found as well?
A translation nobody finds generates no leads.
We do keyword research per language, write native NL and FR copy for the pages that convert, and set up URLs, hreflang and canonicals correctly.
See how our website translation services start from SEO. Stuck on a plugin? Our translation tool migration moves your translations into your own CMS.
Free website scan
Enter your website and get an automatic scan within minutes, with concrete technical and SEO improvements. No sales pitch.
We only use your details for your scan. No spam, unsubscribe anytime.