Wikipedians believed AI would come and go. They were wrong.
In the world of AI overviews, Wikipedia traffic is declining. But the real damage may be far greater than website traffic.
Long before the artificial intelligence storm hit Wikipedia, the Wikimedia Foundation was preparing. Wikipedians broke the hatches and made confident predictions that they would break and wash away. It is said that large language models may even improve Wikipedia, automating housekeeping tasks—formatting, citations, citations, missing links—and thus freeing humans to do creative work. It was not that simple. In 2019, Kathryn Maher, then president of the Wikimedia Foundation, wrote a New York Times article titled "Without Humans, AI Can Wreck." "Too often, artificial intelligence is presented as a powerful solution to our problems, a scalable replacement for people," he wrote. The problem that Wikipedia faced, maintaining human control over machines, was one that it soon faced everywhere.
As early as 2018, Google was training AI to write "Wikipedia-like articles" and by October 2024, a Princeton University study found that about 5% of the approximately 3,000 new articles created on the English Wikipedia in one month (August 2024) contained a significant amount of artificial intelligence generation. There were signs. ChatGPT loved em dashes — parenthetical expressions that look like this — and curly quotation marks. Sometimes, remnants of the notices used to write articles remain (they might say "Here's your Wikipedia article").
Initially, Wikipedians tried to shape and control the AI to make its output "acceptable and acceptable" constructively, or to modify or remove it if proven otherwise. Generic fluff such as "praising a village for its fertile farmland, livestock and scenic countryside" and bogus references were removed. As of March 2026, AI content has been found to violate its core principles so much that writing and translating large language models is banned.
A widely read article entitled "The Death of Wikipedia" explains this defensiveness. This predicted a ripple effect: as contributors reduced their contributions in response to AI infiltrating the platform, there might be a temptation to replace their work with more AI, creating a downward spiral. One Viral X post lamented that AI seems to be taking things in the wrong direction: "I want AI to do my laundry and dishes so I can make art and write, not AI to do my art and writing so I can wash my clothes and dishes." The Death of Wikipedia warned of something similar: instead of AI taking on Wikipedia's "housekeeping" roles, it might write its own articles instead, leaving humans to format and clean up. The result traps editors in an endless curatorial project informed not by human preferences but by the frozen impressions of machines.
Another Wikipedian wrote: "The potential reduction each year of our pool of talented authors is the biggest threat to Wikipedia, because we need to increase the number of editors, especially expert editors, to correct the vast inferno of weak, inaccurate and incomplete articles that are delaying the project." A Wikipedia written and read by machines threatens to create a new, degraded Wikipedia, stripped not only of its context, but of its animating enthusiasm. It will be artless and soulless. When the Wikimedia Foundation says that Wikipedians' care and commitment to "reliable encyclopedic knowledge is something artificial intelligence cannot replace," it doesn't mean it can't be replaced.
When Google introduced AI reviews to search, they quickly became as central to the experience as Wikipedia was until recently. (This also means that Wikipedia still plays an important role as a source, just removed in one step.) Anyone who uses AI agents knows that they give plausible, almost intuitive answers, which can be a superficial check of their source. They are also often wrong. If Wikipedia itself is broken by artificial intelligence, how will we prove them wrong?
If you purchase something using the links in our stories, we may receive a commission. This helps support our journalism. Learn more.
Also inherent in the idea of a single, machine-edited encyclopedia to rule them all is that such a phenomenon should serve as a definitive reference. Its appeal, for those who find it fascinating, is that it will make perfect sense, a kind of digital digital king that can act as a dispassionate arbiter. There is no reason to think otherwise such cold logic would not be cruel. Its aesthetic flaws are a clue: "Slop" is a handy word for an AI product, implying something bland and monotonous that not only lacks nutritional value, but also lacks the flavor that makes eating enjoyable.
There is also no reason to believe that an AI encyclopedia or any other encyclopedia should be unifying. For all its shortcomings and decentralization, Wikipedia is the closest thing we've ever had to an "encyclopedia of everything." When smartphones became widespread after the release of the iPhone in 2007, the old, albeit widespread, speculation came true: Wikipedia was inexpensive, portable, instant, and electric. Vast and all-encompassing, it became truly an approximation of "the sum total of human knowledge". But so what? What did access to this temple mean for human flourishing? What is the use of Wikipedia?
I was asked an exact version of this question by a documentary filmmaker who wanted to show the story of Wikipedia. Yes, Wikipedia was a small encyclopedia that could. It had overcome humble roots to become great. But the TV broadcast commands needed something more. Was there any real-life episode where Wikipedia "saved the village," he asked? Has Wikipedia saved lives? The answer is "probably," though not in dramatic or famous ways. Rapid-fire encyclopedic knowledge has contributed to accidental poisoning, and there are indications that Wikipedia is accessed with particular frequency during natural disasters. But these anecdotes are difficult to verify, and often appear coincidental when Wikipedia's role is examined. If someone calls for help in an emergency, it's not the phone that's the hero.
Does Wikipedia do a public good? This seems obvious, but it's much harder to measure and kind of out of the hands of the project. In 2013, when an online poster posed the question, "If someone from the 1950s suddenly appeared today, what would be the hardest thing to explain to them about life today?" The top answer was broadcast as a bad sign of the times: "I have a device in my pocket that can access all information known to man. I use it to watch pictures of cats and fight strangers." Elsewhere, the "seeing with strangers" line was couched in less light-hearted ways. The academic David Runciman wrote two years later: "The greatest threat to democracy may no longer come from human agency, but from new forms of technology, which have the capacity to undermine democracy without anyone intending to." The forces from which democracy needs to be saved no longer press upon it from without, but from within.
No one foresaw Wikipedia more clearly than H. G. Wells, a reputation built on the collection of articles collected in his book The Global Brain, and its advocacy of a "permanent global encyclopedia." But Wells explained that the idea was much more vague in his other writings, and he even satirized his obsessions in The Camford Visit: “He wants a world of universal education, trained along the same lines, with every artificial difference removed. Shortly after the publication of World Brain, Wells had completely lost faith in his encyclopedic vision and succumbed to late-life pessimism. Humanity had organized to create a world war, not a world brain, and in the civilizational race between "education and disaster," as he called it, disaster seemed to have a winning edge.
It is less obvious to us now that the World Brain was also an antidote. The most important new European encyclopedias before it, the Great Soviet Encyclopedia and the Encyclopaedia Italiana, were direct expressions of nationalist and totalitarian intentions, such that the entry for "fascism" in the Encyclopaedia Italiana was attributed to Benito Mussolini himself. Wells knew that "internationalism" needed his encyclopedia as a counterweight, but he did not make it a simple force for freedom. He envisioned a world brain helping a world government project its power, and though democratically accessible, his permanent world encyclopedia would make such a world government omnipotent. An all-seeing and omniscient world government will inevitably be repressive. The sinister side of Wells' ideas seems to have partly inspired George Orwell's Nineteen Eighty-Four.
Orwell also criticized Wells more bluntly in his 1941 essay "Wells, Hitler, and the World State." One of its more perceptive observations is that Wells was not really a futurist. Orwell thought he was a nostalgic, a "19th-century liberal" who had spent 40 years repeating the same idea: that history is a battle between "scientific man working toward a planned world government" and "reactionary trying to revive a troubled past." With Orwell's help, we can see that Wells' global brain was an old idea, not a new one. Versions of it predated Wales by centuries, and this particular encyclopedic worldview grew out of his Edwardian take on progress with a capital P.
The encyclopedia handed down to us from the Age of Enlightenment is much more than a collection of facts in a book or a website. It is a whole way of seeing and doing, a lens through which to see a clear, scientific view of everything, unencumbered by superstition. Witness the world like this and the world will change. The idea is so widespread that the philosopher Alasdair MacIntyre classified the Encyclopaedia as one of three main approaches to Western ethics, what he called "competing versions of moral inquiry." The others were the "tradition" that began with Thomas Aquinas and Aristotle, and the "genealogy" that began with Nietzsche. The marriage tradition was faith and reason. Genealogy emphasized hidden power relations. The encyclopedia relied only on reason and tried to create an objective picture of the world that could determine how to act thanks to a consensus version of truth. Its purest expression can be found in the 19th century, in the ninth edition of the Encyclopædia Britannica, which promised "the existing facts of human history, collected from the widest regions ... carefully coordinated and grouped, with the hope of the final evolution of the laws of progress, moral and material, which underlie them."
When three competing versions of Ethical Inquiry were published in the 1990s, McIntyre's concept was already criticized for being outdated. One reviewer complained that "the encyclopedias are annoying with their 19th-century dogmatic statements in the Britannica, downplaying the fact that few of their heirs hold such views today." However, there were places where the rational-liberal view remained unapologetically so. It remained at the forefront of the Encyclopaedia Britannica - 1987 edition with the motto of its partner, the University of Chicago: "Let knowledge grow more and more and thereby enrich human life." Most importantly, it was underpinned by the beliefs of Wikipedia's founders. Wikipedia's text-heavy format not only reflected a centuries-old Enlightenment tradition, but was created by two men who still firmly believed that discovering and learning from facts was not just the goal of an encyclopedia, but the central task of a good life.
The divergent paths of Jimmy Wales and Larry Sanger are an allegory for the fate of this worldview and the cross-cultural currents that surround it. Twenty-five years after founding Wikipedia, Wells is still a Davos man, an international person for an international project, living in the British capital and a frequent critic of the MAGA movement. Sanger never left Ohio, where she homeschooled her children, became a vocal vaccine skeptic during the Covid-19 pandemic and voted for Donald Trump. In 2025, he announced that he had abandoned his longtime agnosticism to become a professing Christian and wrote a complete philosophical argument for the existence of God. If Wells had stuck to a McIntyre-style "encyclopedia"—Enlightenment rationalism, with an objectivist term added—Sanger had returned to "tradition."
As Wikipedia asserted its status as a knowledge institution, it lost its immunity to the crises that plague knowledge institutions everywhere. The only surprise was that it had been exempted for so long. At the end of her tenure as president of the Wikimedia Foundation, Sue Gardner likened the experience to living on a peaceful tropical island with "cruise missiles flying overhead." The culture war was raging everywhere and Wikipedia was no longer an island. In the 2010s, "misinformation," "misinformation," and "fake news" were all briefly named buzzwords of the year, with information saturation as much to blame as its corruption. Instead of arbitrating between facts like judges, audiences acted like lawyers who put together a brief and took only what they wanted. The results were "do your own research," an unusual version of the enlightened demand to "dare to know," personalized algorithmic pseudo-information systems. Immersed in this uncertain environment, Wikipedia approached its 25th anniversary with unprecedented political, cultural and technological threats.
All this added a mournful note to his birthday celebration in January 2026. Research firm Pew marked the occasion by quantifying the encyclopedia's staggering reach — in the decade between 2015 and 2025, Wikipedia has been accessed 1.9 trillion times, an average of 508 million hits per day. Nevertheless, these huge numbers included an undeniable arc of decadence. In the days leading up to the anniversary, longtime Wikipedian Christoph Henner issued a lengthy and detailed "wake-up call" to the Wikipedia movement. As of 2022, the encyclopedia had lost nearly a quarter of its readership, more than 1.1 billion hits per month, a lost audience that likely sought AI summaries instead. Even when answers provided by AI agents contain links, readers tend not to follow them: one figure shows that the click-through rate through AI chatbots was only 0.33 percent. For the Google search engine, it was nearly 9 percent. It was a disaster for a model whose lifeblood was editorship.
Artifact warned that AI was an existential threat to Wikipedia, recalling the encyclopedia as a "historical artifact on which AI was once trained ... a footnote in the training data of models that have gone on without us." The editing crisis meant that fewer and fewer editors were doing more, as he put it: "a smaller club working harder to maintain something that fewer people see." Past opportunities—for example, when Wikipedia might have taken advantage of Internet growth outside Western countries—were largely wasted. The energy generated by places like India and Nigeria going digital has overtaken Wikipedia, and in the wake of it all, the encyclopedia is in danger of becoming a "temple", maintained by the last spiritual class of retired Wikipedians. Once offensive ideas, such as paying editors for their work, had to be reconsidered. Otherwise, the whole Wikipedia might turn into an abandoned cathedral.
The media coverage of this occasion was equally bittersweet. Dariusz Gemilniak wrote in Nature that "academic indifference" and artificial intelligence threaten Wikipedia's survival, accusing the agent of exploiting a "parasitic relationship" that "endangers the last bastion of free, human-controlled knowledge. It undermines the very premise of collaboration on which so much of Wikipedia's knowledge-sharing practices rely." "Wikipedia's Existential Threats Are More Felt Than Ever" was the title of a piece by Stephen Harrison in WIRED, listing the threats challenging the project. More dangerous than hostile think tanks and legislators was a society that was increasingly apathetic and slowly drifting away from its fundamental ideals. "Economic instability has trained many generals to view any unpaid work as a fraud," Harrison wrote. In the cold light of the dawning age, becoming a Wikipedia super-editor during a period of unemployment seemed a distant luxury, or downright confusing, "a thankless job that AI companies extract for profit."
The wave that started at Wikipedia was more than the rise of artificial intelligence or the fragmentation of reading habits. Christoph Henner thought that "AI abstractions" were just a "convenient explanation". He wrote: I have doubts:
What we are witnessing is something deeper: a generational shift in the way people relate to knowledge itself. Younger users are not searching. They scroll. They don't read articles. They consume parts. The encyclopedic form factor, our twenty-year-old bet, may be losing relevance faster than any technology can explain. AI is an accelerator, not a fire.
Even this little bit of change undersold. Something much older was also rapidly losing relevance: it was the beginning of the end of the age of the encyclopedia. Near the French front in the encyclopedia there is a diagram called "The Pictorial System of Human Knowledge," now sometimes called "Diderot and D'Alembert's tree." A tree is a system not only for arranging "pieces of knowledge," but for connecting them in a logical chain so long that it could crown the world. It represents a bold idea: that the unifying force of an encyclopedia shows how everything in existence is connected. It shows that knowing everything leads to knowing a great deal about everything. This old idea was revived in the computer age and gained momentum with the concept of AGI, Artificial General Intelligence, the moment when machines surpass any form of human understanding. If realized, it will not only be the end of encyclopedias, but also their climax. Human-authored encyclopedias will still exist post-AGI, but only as secular scriptures in the mind of a digital god.
The oldest criticism of the "encyclopedia" is religious: the belief that human activity on a godlike scale is doomed to failure. It was only during the Enlightenment that absolute knowledge was considered useful (and today, "know it all" is still an insult, not a compliment). Previously, the desire for absolute knowledge was considered a sin or crime, a prohibition that persists in many cultures. Greek mythology is full of cautionary figures like Pandora and Prometheus who knew too much and suffered for it. When the Norse god Odin searches for a sacred well of water that bestows wisdom, a drink costs him an eye. In the first book of the Bible, two different stories, the Tower of Babel and the Garden of Eden, share the same moral twist: full understanding must come from God alone. Curiosity is not only a sin, but the first sin.
One of the many artworks inspired by Wikipedia talks about this phenomenon. In the 2010s, bio-artist Joe Davis announced that he would engineer a real, living tree of knowledge, the tree described in the Bible as bearing forbidden fruit and leading to the fall of mankind. To do this, Davis planned to take a 4,000-year-old variety of apple tree and copy the English language version of Wikipedia into its DNA. He called the project Malus ecclesia, a Latin pun that means both "Church of the Devil" and "Church of the Apple Tree." In interviews, Davis declined to say whether he would eat the fruit from the tree. Due to genetic modification laws, eating an apple or even planning to eat it may be a criminal act. This tempting idea, drowned in red tape, never came to fruition, but its forbidding message remains: the sum of all knowledge is cursed.
The real Wikipedia is less melodramatic than that, and Wikipedia's most enduring benefit may not be its content but its process, which promises quite the opposite of a knowledge tree or even Diderot's tree. For years, the annual Wikimania conference has maintained a fledgling tradition: Wikipedians wear their national costume during the opening ceremony. When almost three-quarters of the countries on Earth are in attendance, it makes for a real spectacle even against the backdrop of the convention center carpet. All around the room are people wearing hanbok, kaftan, and kimono (a practice started by a Japanese Wikipedian), and when they come together, the effect is similar to those old, very hopeful pictures of "people of the world." It takes more time to understand the depth and variety displayed. Those golden scarves called Serimpak are worn by the seafaring tribes of Borneo, which is why they are shaped like ships. The chapan dresses, which seem to spin under their bright colors, are from Central Asia. Those heavy ponchos, too much for air conditioning, come from the Andes. Someone offers Jimmy Wells a conical nomadic hat from Burkina Faso, but he can't accept all the accessories he's offered. It is too much.
They may be a far cry from lederhosen, although someone jokes that Germany's true national costume is sandals with socks. It seems that at least one techie type wears Hawaiian shorts for a similar reason, like Silicon Valley's formal attire. Everywhere there is embroidery, silk and nude colors, scarves and hats, a display of unknown symbols and motifs. People who wear this dress do not create an atmosphere of pride after gathering. In this room their common mission seems humble, their conversation a slow and uncertain work of mutual understanding accompanied by the tentative companionship of shy people. They are limited by human fallibility, which is what gives them dignity. Together they recall a Latin phrase very different from Malus ecclesia. E Pluribus Unum - Out of many, one.