For decades, we have measured progress in digital access largely by asking a simple question: Can people get online?

In the age of artificial intelligence, that question alone is no longer enough. AI is quickly becoming the intermediary between people and the digital world, shaping how we search, summarize, translate, learn, create and communicate. That raises a more urgent question: What happens when these systems do not understand your language?

The Gap Between Connectivity and Language Access

There are more than 7,000 languages spoken around the world, yet roughly ten languages account for most online content. New research from the Coalition on Digital Impact (CODI) illustrates the gap between being connected and being able to meaningfully participate online.

We surveyed organizations working directly with underserved language communities around the world. Only about 29 percent said those communities can navigate the Internet in their native language. Where local-language content exists, it is often concentrated in news and social media, while educational resources, government services, professional tools and commercial platforms remain largely available only in dominant languages.

As AI becomes embedded into education, commerce, public services and everyday communication, that gap matters even more. AI is being built on top of a digital ecosystem that has never worked equally well across languages. Without intervention, it risks magnifying those inequalities into the next generation of technology.

How Language Barriers Reinforce Digital Inequality

The barriers can appear deceptively small: a keyboard that does not support a language. An online form that rejects an internationalized email address. A search engine that fails to surface local-language content. An automated translation that misses cultural context badly enough that people stop trusting it.

Taken together, those failures create a digital ecosystem in which millions of people must repeatedly leave their language behind to participate online. Even then, they’re only able to participate if they speak a dominant language, and the depth of their online experience is dependent on their level of fluency.

When a language is poorly represented online, there is less content and data available for AI systems to learn from. If those systems then perform poorly in that language, people have fewer useful tools for creating content, finding information and participating online. If investment continues to flow primarily toward the largest markets and datasets, dominant languages can become even more entrenched, and the digital divide becomes self-reinforcing.

What Meaningful Language Access Requires

Meaningful language “access” requires more than a language existing somewhere online; it must work across the entire digital experience. At CODI,we have mapped out exactly where language must be supported across the Internet to enable meaningful participation, and the framework is revealing.

People need to be able to type and communicate in their language. Their domains and email addresses need to function. Relevant content needs to exist and be discoverable. Search and AI systems need to understand it. Translation needs to be accurate and culturally grounded. And amid elevated AI risks, critical information like security warnings and fraud alerts needs to reach people in the language they actually understand.

Bridging the Investment and Implementation Gap

There is already significant investment around the world in broadband, digital skills, governance and technological innovation. But our research found a persistent blind spot: linguistic inclusion is rarely treated as a foundational design requirement. Too often, language is added later through translation or localization after the underlying system has already been built.

We also found a disconnect between the people closest to this problem and those with the resources to address it. Last-mile organizations told us they need sustainable funding, better language technologies, stronger data, localized content and technical capacity. While governments, technology companies, philanthropies and international organizations are investing heavily in digital development, these two ecosystems often operate in parallel.

AI gives us an opportunity to build that bridge. We must involve local language communities in technological development, invest in foundational language data and tools, implement Universal Acceptance so that domains, email addresses and scripts work across digital systems, and treat linguistic inclusion as critical core infrastructure in the digital world.

The Internet connected the world without fully solving the language problem. AI now raises the cost of leaving that work unfinished. The choices we make today will determine whether AI reinforces the linguistic divides of the Internet we inherited or helps us finally begin to close them.