Google Is Already Translating Your Competitors’ Content: You’re Probably Next

Our research into Google’s AI Overview reveals something unsettling: when your site isn’t translated, Google does it for you. Under its own domain. Without asking.

Here’s something most international businesses haven’t caught onto yet. Google isn’t waiting for you to translate your website. It’s doing it anyway – serving up its own translated versions of your competitors’ pages to foreign-language searchers, under Google proxy URLs, with zero input from the original brand.

We ran a study to figure out exactly how often this happens, and the results were more aggressive than we expected.

Why we started looking into this

ai overview study

We’ve spent the past year studying how AI-powered search handles multilingual content. Our previous research analyzed over 1.5 million citations across Google AI Overviews and ChatGPT. The conclusion was hard to argue with: if your website exists in only one language, AI search treats you like you don’t exist in any other.

The data backed that up convincingly. Sites that added even one translation saw visibility jump by 291%. Translated sites earned 31% more citations per query on average, including a bump in their original language. There’s something about multilingual presence that makes AI systems treat your content as more authoritative.

But that raised an obvious follow-up question. If Google prefers translated content, what happens when it can’t find any? Does it just leave a gap in the results? Or does it take matters into its own hands?

What we tested

We wanted to know: when content doesn’t exist in a searcher’s language, does Google translate English sources itself and serve them through proxy addresses?

We weren’t the first to notice this behavior. Ahrefs flagged it initially—they found Google displaying translated pages under its own proxy domains rather than linking to the brand’s actual site. That research is what put us on this trail.

The tricky part was isolating language as the variable. Certain query types—”where to buy,” “near me,” anything with local intent—trigger map results and business listings regardless of what language you’re searching in. We had to design around that.

Study design

We built a controlled test: 75 queries in English and 75 identical queries translated into Russian, all executed from the United States.

Russian was the right choice for a few reasons. It’s not a dominant language in US search, the Cyrillic alphabet makes it easy to identify translated-vs-native content, and Google’s AI Overview is active in Russia. These conditions gave us a clean way to measure what happens when content in a specific language is limited.

We excluded any query with location intent and focused purely on informational and commercial searches – the kind where website content should be the answer, not a map pin.

Across 150 queries, we examined 1,263 cited sources.

The results

In Russian-language results, 11.4% of all cited sources were English pages that Google had translated on its own. In the English results? Zero. Absolutely none.

Google is only doing this in one direction: from English into languages where it can’t find enough native content.

English queries got dramatically more sources

AI Overview cited an average of 11.3 sources per query in English versus just 5.8 per query in Russian. That’s nearly twice as many sources for English-language searches.

When Google can find content in the query language, it uses it. When it can’t – especially for Russian – it fills the gap with its own translations. But the total number of sources still drops significantly, which means even with Google’s intervention, non-English searchers get a thinner experience.

You’d rather be the native-language result than the one Google awkwardly proxied. And your customers would rather land on your actual site.

The industry breakdown tells a clear story

Not every sector was equally affected:

Nearly a quarter of all tech sources in Russian results were Google translations of English pages. That’s enormous. SaaS wasn’t far behind. These are industries where English-language content dominates and local alternatives are sparse—exactly the conditions where Google steps in.

Retail was lower because local shopping content tends to exist in most languages. B2B barely registered, though it wasn’t zero—even in sectors with decent Russian coverage, Google still occasionally pulled in translated English pages.

The pattern is clear: the less content that exists in a given language for your industry, the more aggressively Google fills the void with its own translations.

What this means for your business

We think this data points to something bigger than a search quirk.

If you haven’t translated your site, you may already be visible in foreign markets—just not on your terms. Google’s proxy translations mean a Russian-speaking user could be reading your content right now, on a Google URL, in a translation you never approved, with no link to your actual domain. Your brand name might appear in the text, but the user never visits your site, never sees your design, never enters your funnel.

You’re present in the search results. You’re just not there.

Translation has become a control issue, not just an access issue. A year ago, the argument for translating your site was about reaching new audiences. That’s still true, but now there’s a defensive angle too. If you don’t translate your content, Google might do it for you—badly, without context, and under a URL you don’t own. Every page Google proxies is a page where you’ve lost control of your messaging, your brand voice, and your traffic.

The gap is getting filled whether you act or not. This was a focused study—150 queries, four industries. But it lines up with everything else we’ve seen in this space. AI search tools are not shy about rewriting, translating, and repurposing content to serve users in their preferred language. Google’s AI Overview doesn’t shrug when it can’t find Russian-language content. It manufactures it.

Somebody is going to translate your site for your target markets. You can do it yourself—on your domain, in your voice, with your brand front and center. Or you can let Google handle it, and hope for the best.

Methodology

We analyzed 150 queries (75 English, 75 Russian) across four business categories—SaaS, B2B, Retail, and Tech—in Google’s AI Overview. All searches were conducted from the United States. Queries were designed to exclude location-based triggers and isolate language availability as the primary variable. A total of 1,263 sources were examined for translation status and origin.

ConveyThis translates your website automatically and puts you in full control of every language version. No proxies. No guesswork. No handing your brand voice to an algorithm. Try it free.