Wikipedia is important.

How big it actually is

Wikipedia is the fourth most visited website on earth. Not the fourth most visited encyclopedia. The fourth most visited website, behind Google, YouTube and Facebook, and ahead of Instagram.

Those last two numbers are the interesting pair. Sixty-six million articles, and only about six hundred thousand people who edited anything at all in the last month. Across the entire planet, the number of people actively writing the world's reference work would not fill a mid-sized city.

And then there is the position it holds. When someone types a question into a search engine, Wikipedia is very often the first result, or the box of text above the results. People are not usually choosing Wikipedia. They are choosing to look something up, and Wikipedia is what looking something up gives you.

Why what is written there changes what people do

Information is only interesting if it touches a decision. This one does, constantly, in small ways that add up.

Take food, because it is the thing I think about. What a person eats has consequences for animals, for the environment, for their own health, and for the health of the community they live in. It is a decision made several times a day by everyone. And when someone goes looking into a particular way of eating, what they find is very often a Wikipedia article.

Consider the article on the Atkins diet, and the article on the man who created it. A reader arriving there does not find a verdict. They find the documented record: that Robert Atkins died in 2003 at seventy-two of a head injury after falling on ice, and also that he had suffered a cardiac arrest the year before, which he attributed to an infection-related cardiomyopathy rather than to his diet, and also that a medical examiner's report noted a history of heart attack, congestive heart failure and hypertension, and also that his family and his critics have argued about what all of this means ever since.

That is a messier story than either side would like. It is also exactly the point. Nobody gets to write the tidy version. Both the claim and the objection are on the page, sourced, permanently, and a reader can see the shape of the dispute for themselves. That is worth more to a person making a real decision than any confident sentence would be.

And it is not only members of the public reading. It is journalists checking a background before filing. It is policy staff getting up to speed on a subject before a briefing. It is teachers, and it is children, who will go on being adults for a long time.

The part that is new

Large language models are trained on Wikipedia. Not exclusively, and not even mostly by volume, but heavily and deliberately, because it is one of the largest bodies of clean, structured, openly licensed prose that exists. When the systems that hundreds of millions of people now ask questions of were being built, Wikipedia was in the room.

They also reach for it while answering. Across studies of which sources AI assistants actually cite, Wikipedia comes out as the single most-cited domain on the web. The exact share moves around a great deal depending on who is measuring and when, and I would not put much weight on any particular percentage. The ranking is the stable finding, and the ranking is the part that matters.

For the same reason that being at the top of a search result mattered to policymakers, to businesses, and to anyone working in education, being in Wikipedia now matters to the models that are already shaping what those people believe.

So the leverage has changed shape. It used to be that an edit reached the people who read the article. Now an edit also reaches, indirectly and imperfectly, everyone who asks a machine about the subject and never visits the article at all. Putting good information on Wikipedia has become a way of telling these systems that this information is true, and that it is important.

There is a real complication worth naming, because I would rather say it than have you find it. In late 2025 the Wikimedia Foundation reported that human page views were down roughly eight per cent year on year, and attributed the drop to AI summaries answering questions that people used to click through for. Fewer humans are arriving at the article. That is a genuine loss, and it is a problem for Wikipedia's volunteer pipeline and its fundraising.

But it cuts the other way for the argument I am making here. The readership is being routed through the models rather than the page. The information still arrives. It just arrives with the URL stripped off. Which means the content matters as much as it ever did, and the credit matters less.

Why this is a lever you can actually pull

Most of the ways a person can help with anything large are frustrating in the same way: the connection between what you did and what changed is invisible, and you have to take it on faith.

This one is not like that. I have started calling it a centre-of-the-board move. It is not a bet that pays out or doesn't. It is straightforwardly proportional. If you make ten edits, you get the result of ten edits. If you make twenty, you get the result of twenty. Whatever you pour in has a result of roughly that size, and it stays there, accumulating views, at no ongoing cost to anyone.

It is also an on-ramp, and I think that is the more important half. Every movement working to end some kind of suffering depends on the hope that more people will put more in. And there are orders of magnitude more people who might help than are currently helping, in any cause you care to name. What those people usually lack is not conviction. It is a first thing to do that is small enough to actually start.

Editing works for that. It can be done from a couch, alone, at any hour, by someone who is not comfortable in a crowd, in whatever amount of time they have. Wikipedia has already demonstrated, over twenty-five years, that it can hold the attention of volunteers who are paid nothing.

And every action a person takes towards the world they want is partly a vote about who they are. Small steps matter disproportionately at the beginning, because they are what turns someone who cares into someone who does things.

Doing it properly

Wikipedia has rules about this, they are enforced, and they are worth understanding before you start, because getting this wrong sets a cause back rather than forward.

The core policy is neutral point of view. Related to it are the policies on advocacy, which is defined as using Wikipedia to promote your beliefs at the expense of its goals, and on conflict of interest, and on coordinating groups of people to sway a page, and on paid editing, which requires disclosure and is not permitted for advocacy at all.

What this rules out is a campaign to make articles say what we would like them to say. What it does not rule out, at all, is a campaign to make articles better: more complete, better sourced, less shaped by whoever happened to be paying attention.

There is good precedent for organised effort inside those lines. WikiProject Women in Red was started in 2015, when only 15.5% of English Wikipedia's biographies were of women. It is a coordinated, openly declared, deliberately targeted campaign, and it is entirely uncontroversial, because its method is writing well-sourced articles about people who merit them. By 2026 the figure had passed 20%, on more than two million biographies, with over 200,000 new articles written. There is a companion project on women's health doing the same for medical coverage.

That is the model. Not louder opinions. Missing articles, written properly.

The smallest useful thing

The biggest barrier is not willingness. It is that people genuinely do not know how, and the interface does not obviously invite them. So here is a ladder, and the first rung is very low.

1. Fix something small.A typo, a broken link, a sentence that does not parse. You do not need an account. There is an edit button. This exists mostly to prove to yourself that the button works.
2. Make an account.Takes a minute, gives you a watchlist and a talk page, and makes your work attributable to a person rather than an address.
3. Add a citation to a claim that has none.Look for the "citation needed" tags. Find the source. Add it. This is the single most useful low-skill contribution on the entire site, and there is an effectively unlimited supply of it.
4. Add a sourced sentence to an article you know something about.Something true, relevant, and referenced to a source someone else could check.
5. Adopt a thin article.Find something in your field covered in three sentences that deserves thirty. Spend a few evenings on it. This is where the real value is.
6. Write one that does not exist.Hardest rung, most rewarding, and the one where the notability rules matter most. Read those first.

Rungs one to three can be done today, in less time than it takes to read this letter twice. If you only ever do rung three, repeatedly, you will have done more for the world's shared reference work than almost anyone you know.

If you teach

If you are an academic, or a teacher of any kind, you have access to something the rest of us do not: a room of people who are already being asked to research a subject and write about it carefully, for an audience of one, after which the work is filed and never read again.

That last part has always seemed like a waste to me. The same effort, pointed at Wikipedia, produces something that stays.

The argument I would make to a colleague is this. The reason academics write papers at all is to share findings and contribute to a field's discourse. The difference between a hermit running experiments and a scholar is not the experiments. It is the sharing. In the nineteenth century the best available routes were journals, conferences and letters. Those still matter. But dissemination now runs substantially through the encyclopedia that sits at the top of every search and inside every model, and a field that is thinly represented there is a field that is quietly absent from what the public and the machines both believe.

So the ask is not "invent a Wikipedia assignment." It is "adopt one that has been refined over a decade, and add your students to a pipeline that is already the largest single source of new editors on the site." Outside North America, the wider Wikipedia Education Program does the same work.

None of this is a grand gesture. That is the appeal. It is a small, boring, permanent, compounding thing that almost nobody does, sitting at the front of how a very large number of people, and now a very large number of machines, come to believe what they believe.

Truth is on our side in most of the arguments I care about. It just has to be written down, in the place people look.

— Em

Where these numbers come from

Every figure on this page is linked to its source. Where the evidence is contested or the number moves around, I have said so in the text rather than picking the flattering version.