Google Is Translating Your Competitors' Content (And It Could Be Yours Next)

A small-scale study offers early insights into how Google's AI Overview might be handling the translation gap while raising questions about traffic control in multilingual search.

As search evolves into something more unpredictable and AI-driven, website owners are grappling with yet another new challenge: maintaining control over their traffic in an increasingly multilingual digital landscape.

We wanted to understand what happens when Google's AI Overview encounters queries in languages where local content is scarce. Does it simply skip those websites? Or does it take matters into its own hands?

Turns out, it's the latter. And businesses with international ambitions will want to pay attention.

We've been researching how AI search handles multilingual content. Our previous study revealed that untranslated websites don't get cited in AI search. Here's the video if you want a quick summary:

This study takes the next logical step: what happens when Google needs content that doesn't exist in a user's language?

What We Wanted to Find Out

There was a problem with our initial hypothesis.

When we first set out to study Google's proxy behavior, we wanted to answer a straightforward question: When translated content isn't available in a user's language, does Google translate English sources itself using proxy addresses?

In fact, it was the team at Ahrefs that first signalled this anomaly: they found that Google frequently displayed translated content under its own proxy domains rather than directing users to the actual brand websites. That gave us the inspiration for our previous studies on AI search visibility, as we suspected this behavior was connected to the availability (or lack thereof) of properly translated content.

But our research flagged an important complication: the appearance of Google proxies also seems to be linked to the query itself. When people search for things like "where to buy [product] in person," Google's algorithm prefers to show locations, phone numbers, and reviews rather than website links, regardless of whether the content is translated.

This meant we couldn't just run searches in different languages and compare proxy rates. We needed to design our test carefully to isolate the language variable from the query type variable.

The Methodology

To isolate the language variable from query type, we designed a controlled experiment comparing search results in two languages:

• Control group: 60 queries in English, performed in the US
• Test group: 60 identical queries translated into Polish, also performed in the US

Why Polish? Two reasons: it's not widely used in the United States, and AI Overview is available in Poland, making it an ideal candidate for testing language availability.

This is where query design became critical. We generated 60 business-oriented queries specifically designed to avoid location-based triggers, like none of the 'where to buy' or 'near me' searches that would naturally prioritize Google Maps or local listings regardless of language. Instead, we focused on informational and commercial queries where website content would be the primary answer.

In total, we examined 1,048 cited sources across 120 queries.

The Results: Google Steps In When Translation Is Missing

The data offers an interesting glimpse.

In Polish search results, 7.9% of sources were English websites translated by Google. But in English search results, it fell to zero. Not a single source was translated.

But that's just the headline figure. Let’s look at how Google's AI Overview behaves differently across languages.

English Gets Way More Sources

On average, AI Overview cited:

• 10.72 sources per query in English
• 6.75 sources per query in Polish

That's 58% more sources for English queries.

This suggests that Google's AI Overview prioritizes content in the query language when it's available. When it can't find enough Polish content to provide a comprehensive answer, it fills the gap by translating English sources. Still, it clearly prefers native language content when possible—another argument to translate your website (and own all the benefits, rather than being at the mercy of changing search landscapes).

Not All Industries Are Affected Equally

When we broke down the data by industry category, we found that some sectors are far more likely to be translated by Google than others:

Tech content was more than twice as likely to be translated compared to the overall average. The reason makes sense: tech queries are highly specific and often use English-dominated terminology. This pattern suggests Google may be filling content gaps by translating English sources when Polish alternatives are scarce.

Retail content, meanwhile, tends to have more local equivalents. As for B2B, it turns out that there was enough Polish B2B content available that Google didn't need to translate any English sources at all. So the availability of content in a particular language will definitely tip the scales.

What This Means for Your Business

If you're running an international business, this early research points to three trends worth monitoring:

1. You May Be Competing Against Google's Translations

When your website isn't translated, you're not just invisible in foreign markets. You might actually appear in search results... but not on your terms. Our smalllimited sample suggests that Google will translate your content and display it under a proxy address, which means:

• Users don't see your actual domain
• You lose brand visibility and recognition
• Traffic goes through Google's proxy instead of directly to you
• You have no control over how your content is translated or presented

2. Translation Is About Control, Not Just Access

We can all agree that search is becoming more unpredictable. AI-powered results are reshaping how people discover content. In this environment, maintaining control over how you’re found is more important than ever.

When you translate your own content:

• Your actual domain appears in results
• You control the quality and accuracy of translations
• You build brand recognition in local markets
• Users land directly on your site, not a Google proxy
• You can optimize for local search behavior and cultural nuances

Our study suggests that Google may fill the gap if you don't. But wouldn't you rather control your own narrative?

While limited in scope, this research adds evidence to what many have observed: AI-powered search tools are willing to take significant liberties with content to provide answers in users' preferred languages. Google's AI Overview doesn't just ignore non-English markets or stick to English-only results. It actively translates content when local alternatives are scarce.

If this pattern holds at scale, the translation gap may already be getting filled—either by you or by Google.

Methodology Note

This study analyzed 120 queries (60 in English, 60 in Polish) across four business categories (SaaS, B2B, Retail, Tech) in Google's AI Overview. Queries were designed to avoid location-based triggers and focus on informational and commercial intent. All searches were conducted from the United States. A total of 1,048 sources were analyzed for translation status and origin. Queries were designed to avoid location-based triggers and focus on informational and commercial intent to isolate language availability as the primary variable.

Want to see how Weglot can help you maintain control of your international traffic? Try it for free for 14 days, no commitment.