Does AI speak your local language?

The short answer is – it depends on what your local language is.

Picture a European startup that has spent years building something genuinely innovative. A FinTech platform that helps mid-sized companies manage cross-border payments with happy customers at home and across a couple of neighbouring markets.

The ambition, now, is the UK or the US market. The traditional next steps are to translate the website, run some ads, and start reaching out to partners abroad.

But the finance director in London or New York who might buy this software isn’t opening Google and scrolling through links anymore. They open their AI assistant and type: “Which providers in Europe offer the most efficient cross-border payment platforms for mid-sized companies?”

The startup probably doesn’t show up in the answer. Not because the product is weak, but because, to a global AI system working in English, the company is barely visible.

People are asking instead of searching

ChatGPT now has more than 900 million weekly active users, according to OpenAI – more than double the 400 million reported a year earlier. Whilst it’s just one tool’s growth figure, it points to the fact that more people are skipping the list of links and just asking the question, expecting a finished answer back.

Forrester’s State Of Business Buying 2026 report surveyed almost 18,000 B2B buyers worldwide and found that 94% had used generative AI in their most recent purchase decision, up from 89% the year before.

Buyers rated generative AI or conversational search as their single most valuable source of information, twice as often as any other source. More specifically, 55% compare vendors directly inside an AI tool, 54% do product research there, and 47% use it to prepare the case for an internal decision.

All of that happens before a sales team hears a word. Whether your company makes the shortlist is increasingly decided before a single prospect visits your website.

Search now ends without a click

Alongside the rise of AI chatbots, Google search itself is changing. You’ve probably noticed this in your own behaviour over the past year – you ask something simple, like opening hours or an exchange rate, and the answer appears above the results, and the page that used to hold that answer doesn’t get visited at all.

The same pattern is moving into more complex buying decisions. Instead of comparing five open tabs, we ask one question and accept the synthesised answer as good enough.

For a company, this means a website visit is no longer the automatic first step. The website used to be where the brand’s story began, but now much of the decision happens before anyone gets there.

Why a translated website isn’t enough

Traditional SEO was built on keywords and backlinks. Generative Engine Optimisation (GEO) runs on something different. Most AI assistants today aren’t limited to what they learned during training. They search and evaluate live sources in real time to build an answer.

Visibility gets decided when someone asks a question in English and the system reaches first for English-language content it judges credible – foreign media, trade publications, substantial discussion threads. If the only mentions, interviews or case studies about your company exist in your local language, they’re far less likely to make it into the sources the model draws on for an English answer.

In other words, the system doesn’t build trust only from what you say about yourself on your own site. It builds trust from the footprint you leave across the wider digital space, in the language the question was asked in foreign media, in professional communities, in independent sources.

The language gap we don’t talk about

Several independent academic studies of large language models point to the fact that languages with a smaller volume of digitally available text tend to get lower model performance and accuracy than large languages like English.

Last year, Microsoft and the non-profit Common Crawl launched a programme to boost digital content for the languages least represented online. According to the published data, even much larger European languages like Danish, Finnish and Swedish make up less than 0.6% of web content.

That’s not a reason to abandon your own language, but a reason to stop relying on it exclusively if your target customer is abroad. Without a reasonably dense English-language footprint, the odds of a global AI system knowing you exist at all drop significantly.

How to build international visibility

Exactly how each AI system selects and weighs its sources continues to shift. However, four areas are particularly worth prioritising for international visibility:

Try this over the next few weeks and put three questions your ideal foreign customer might ask into a few different AI tools, in English, phrased a few different ways.

Answers shift depending on exact wording, the tool used, and even the day, so one attempt isn’t a diagnosis. But if your company keeps failing to show up, that’s a clear signal of where to start.