Literacy Now Means Knowing What to Trust Online

Key Takeaways

• Literacy used to mean decoding text. It now includes finding information, judging whether a source is reliable, and recognizing manipulation.
• That expanded definition assumes reliable information exists in your language and can be found. For most of the world's languages, neither holds.
• Search engines, AI assistants, and fact-checking resources all perform worse outside a small set of dominant languages, so the same skill produces different results depending on who uses it.
• Digital literacy is framed as a personal responsibility. Much of what determines whether it works is infrastructure nobody is being held accountable for.

Somewhere along the way, the definition of literacy changed, and nobody put it to a vote.

For most of the last century, being literate meant you could read and write. That was the measure, and it was a reasonable one. Now, the same word carries a much heavier load.

Being literate is supposed to mean you can find information, tell a credible source from a fabricated one, notice when something is deliberately engineered to make you angry, and understand that a chatbot's confident tone doesn’t actually mean it's correct.

That's a substantially harder skill set. It's also one that only works if the environment cooperates, and for most people on the planet, it doesn't. This week, September 8th, UNESCO marked International Literacy Day, making this a reasonable moment to look at what the word is being asked to cover.

What Literacy Means Now

The practical definition has expanded to include at least four separate abilities.

Finding information at all, which means knowing how to search and understanding that results are ranked by systems with their own commercial priorities. Judging reliability, which means telling a news outlet from a content farm, or a study from a press release about a study.

Recognizing manipulation, which covers emotionally engineered headlines, coordinated smear or fearmongering campaigns, and synthetic media. And understanding the machinery well enough to know that a feed is curated, that an AI assistant can be fluently wrong, and that neither is a neutral window onto the world.

None of that was part of the traditional literacy conversation. All of it is now required to hold a job, follow a news story, or make a decision about your own health.

Here's the assumption buried underneath all four. Each one takes for granted that reliable information exists somewhere, in a language you read, and can be surfaced when you go looking. Remove that assumption and the skills have nothing to act on.

The Infrastructure Underneath the Skill

There are roughly 7,000 languages spoken in the world. The internet meaningfully supports a small fraction of them.

English accounts for a disproportionate share of all web content while being the first language of a small percentage of the world's population. A handful of other languages cover most of what remains. Everything else divides what's left over, which in practice means thin search results, few authoritative sources, and often no fact-checking organization at all.

UNESCO's Global Roadmap for Multilingualism in the Digital Era treats this as a design failure rather than a natural outcome, calling for digital content, platforms, and tools built to genuinely include the world's languages. That framing locates the problem in the systems, not in the people struggling to use them.

Search shows the gap most clearly. A search engine can only rank what has been published. Where little authoritative material exists in a language, results come back thin, outdated, or dominated by whoever happened to publish something. The person searching did nothing wrong. There's simply nothing good to return.

AI assistants inherit the same gap and hide it better. A model trained mostly on English produces more accurate and more nuanced output in English. In a language with little training data, it produces answers that sound exactly as confident and are considerably less reliable. Nothing in the interface signals the difference.

The Person Who Does Everything Right

This is where framing literacy as a personal skill falls apart.

Picture someone following every piece of digital literacy advice correctly. They check the source. They look for a second confirmation. They're suspicious of anything that arrives with an emotional charge attached. They know an AI assistant can hallucinate a citation or a source.

Now, put that person in a language with a few thousand indexed pages, no local fact-checking organization, and an AI assistant that hallucinates more often than it does in English. Their skills are intact and irrelevant. The environment those skills were designed for doesn't exist where they live.

The people in that position tend to be the ones with the most riding on the outcome. Communities where reliable health information is already scarce. Regions where local journalism has been defunded or suppressed. Places where content moderation fails because nobody at the platform reads the language, so harmful material circulates freely while ordinary speech gets flagged by mistake.

A verification skill is only as useful as the material available to verify against. Handing someone a method with nothing to apply it to isn't empowerment. It's a way of transferring responsibility.

Half the Problem Gets All the Attention

I think the honest version of digital literacy has two halves, and public conversation covers roughly one.

The first half is the skill. Teach people to search well, evaluate sources, and question what lands in front of them. That work is genuinely valuable and worth funding.

The second half is the condition. Make sure something reliable exists to find. That means investment in local-language content, support for journalism in underserved languages, multilingual design built into platforms and AI tools from the start rather than bolted on later, and search systems that don't quietly rank whole linguistic communities into irrelevance.

Only the first half is discussed as an obligation. The second gets treated as a market outcome that simply happened, with no one in particular answerable for it.

That asymmetry is convenient for everyone except the people living with the consequences. It's much cheaper to run a media literacy campaign than to build a functioning information ecosystem in a language with no commercial value attached to it.

Being able to find and trust information is close to a precondition for everything else the internet is supposed to enable. Treating it as a personal aptitude, while the infrastructure that determines whether it works goes unbuilt, gets the responsibility backwards. The skill is worth teaching. The environment is worth building.