Steven Pruitt spends his evenings identifying errors that most people never notice and making fixes that hardly anyone ever thanks him for. He toils at a desk in a town house in Alexandria, Virginia, surrounded by books—the kind of working clutter that suggests a long relationship with paper rather than a fetish for screens. And yet his work is necessarily digital; after dinner, and sometimes late into the night, he uses his desktop computer to correct dates, clean up syntax, standardize categories, and occasionally write entire biographies of people on Wikipedia, the free online encyclopedia.

Wikipedia is not his employer, of course. Like all editors on the site, Pruitt is a volunteer. “At this point, I won’t say I don’t have any skin in the game,” he told me. “But it’s a lot lower stakes than a job because if I get something wrong, it’s fairly easy to fix it. I can fix it myself. I can do what I want to do on my own time.” Still, he holds himself to some rules: “I do try to get in at least one edit a day.”

Pruitt is forty-two, and works full time as a records-management contractor for the federal government. After graduating from the College of William & Mary, in 2006, he moved back in with his parents, owing to the cost of real estate in Alexandria. In recent years, he helped his mother care for his father. (While I was reporting this story, Pruitt’s father died.) In effect, Pruitt—the person who has done more than anyone else to shape the English-language Wikipedia—lives a life that is, by most outward measures, unremarkable.

According to public tallies, Pruitt has made more than six million edits to Wikipedia and created more than thirty thousand articles. He is widely regarded as the most prolific Wikipedian in the entire world. (“It depends on how you’re counting,” he said. “Different tools count different things.”) In 2017, Time magazine included him on its list of the most influential people on the internet. But outside of a small circle of editors, researchers, and obsessive readers on Wikipedia, the recognition has barely registered.

On the site, he is known by his username, Ser Amantio di Nicolao—a reference to a minor character in “Gianni Schicchi,” Giacomo Puccini’s comic opera. Pruitt’s interest in opera is genuine, but the flourish is misleading. He avoids drama, which means that he avoids writing Wikipedia biographies of people who are still alive, whenever possible. “I generally don’t do a lot in the realm of current events,” he told me. “Not just because the stakes are too high but sometimes because there’s so much editing going on on a subject in a particular moment that it can take me five or ten minutes just to break in with one edit.” He prefers biographies of what he calls “fairly obscure dead people.”

“They’re settled,” he explained.

In 2001, Jimmy Wales, an internet entrepreneur, and Larry Sanger, a philosopher, launched Wikipedia as an experiment in collaborative knowledge creation, allowing anyone with an internet connection to contribute. Pruitt first encountered the site in 2003, when he was still in college. “I didn’t quite understand what it was,” he recalled.

For more than a year, he did not edit at all. He would stumble upon Wikipedia pages through search results or links, and then move on. Between late 2004 and early 2005, though, his relationship to the site began to change. The encyclopedia had reached what he described as a critical mass: “There was enough stuff on the site that there was always something to do,” he said. “But it wasn’t just a blank slate.” The difference mattered. A completely empty encyclopedia was intimidating; a partially filled one invited correction and expansion.

Some of Pruitt’s first contributions were related to niche aspects of the Revolutionary War, including an entry on one of his own ancestors, a relatively obscure soldier from Virginia. He wrote several articles on topics and figures—such as Francis Salvador, the first Jew to be killed in the Revolutionary War—that would have been unlikely to appear in a traditional encyclopedia, and he watched, with a kind of amazement, as they remained, becoming permanent parts of the internet. “At this point, it’s over twenty years,” he said.

Pruitt was actively editing during Wikipedia’s first major growth spurt. In 2006, its English-language edition reached a million articles, a milestone that signalled both scale and legitimacy. (Wikipedia now exists in an estimated three hundred and forty-six languages, with the English edition containing the most entries.) As the site became bigger and more prominent, expectations hardened. Editing became less about authorship and more about stewardship. “Even if an article looks complete, honestly, it isn’t ever really complete,” Pruitt said. Every page existed in a state of provisionality, open to revision and dispute.

Over time, Pruitt’s work shifted away from creating new articles and toward improving existing ones. He standardized categories, repaired references, and corrected small errors before they crystallized into misinformation. He has still seen some of those errors migrate elsewhere on the internet. “You fix something,” he said, “and then you’ll see it show up years later somewhere else.”

Today, much of Pruitt’s work is made possible by a specialized editing tool called AutoWikiBrowser. It allows experienced users to make rapid, semi-automated edits across large numbers of articles, flagging common problems and applying standardized fixes. Pruitt uses it to revise at a scale that wouldn’t be possible by analog means: correcting dates, repairing broken links, standardizing citations, and cleaning up formatting across thousands of pages. Despite the name, “it’s not automatic,” he said. “You still have to review every edit.” The tool does not write articles or make judgments. It accelerates maintenance.

As I perched next to Pruitt in his den, he turned to the computer and began doing what he calls “mass editing.” He opened up a web-based external tool known as PetScan, which generates lists of Wikipedia articles by category, allowing him to layer categories the way a researcher might combine search terms. “Women scientists” crossed with “19th-century physicists,” for example, produces a list of biographies that fit both criteria, revealing pages that might be missing a relevant classification.

The method depends on intersections. Categories are layered together until they produce a set of pages that can be worked through in sequence. Too broad, and the list becomes unusable; too narrow, and it returns almost nothing. After Pruitt ran the query for “19th-century women physicists,” a list of articles appeared. From there, he moved quickly. He opened a biography and checked whether it had been categorized correctly. In some cases, an article about a nineteenth-century woman physicist had been tagged as a scientist but not as a physicist. Many entries required only small corrections; others needed more substantive changes. The work consisted of small decisions made in rapid succession, each one nudging the encyclopedia closer to order.

Wikipedia’s hierarchy is deliberately difficult to see. Editors work under pseudonyms. Articles appear collectively authored. There are no bylines, no salaries, no masthead. Pruitt was granted administrative privileges, after another editor nominated him through Wikipedia’s standard Request for Adminship process, where the editing community supported his candidacy. These privileges allow him to block users and close discussions, but he is careful about what that power does and does not mean. Wikipedia discourages editors from reverting—“undoing”—one another more than three times, regardless of correctness. “You can be blocked for twenty-four hours,” he said. “It doesn’t matter if you’re right.” He likes the rule. “It keeps things from turning personal.”

Inside Wikipedia, reputation accrues through time rather than visibility, and it “comes as much from longevity as anything else,” Pruitt said. “You stick around. People know you.”

Among the people who stick around—and who make consistent contributions to the site—are the Wiki-obsessives known colloquially as “super editors.” These individuals are responsible for hundreds of thousands, if not millions, of edits. Although there are more than a hundred and thirty-two million registered accounts on Wikipedia, a study found that one per cent of these users are responsible for roughly eighty per cent of the site’s content.

Justin Knapp—a forty-three-year-old in Indiana who has held all sorts of jobs, including working at a pizzeria and at a crisis hotline—is one super editor. In 2012, he became the first person to surpass a million edits on the English-language Wikipedia, before Pruitt surpassed him in total edit count a few years later. Knapp told me that he doesn’t see Pruitt as a rival. “It’s not competition,” he said. “We’re collaborators. We’re colleagues.” Asked whether editors are motivated by metrics at all, Knapp was unequivocal: “I really don’t think the kind of person who sets about to spend his free time writing an encyclopedia is really the kind of person who cares a lot about being a big fish in a small pond.”

That ethos—low ego, high volume—helps explain how the site functions. Both Pruitt and Knapp described Wikipedia less as a platform than as an ongoing negotiation among strangers. Editors coördinate through talk pages—discussion threads attached to every article, some of which run for years. Disputes accumulate on talk pages over phrasing, sourcing, whether a subject belongs at all, with many never resolving. A finished Wikipedia page reflects a momentary consensus, subject to revision as soon as someone decides to revisit it.

A Wikipedia page’s internal debates often center on matters far removed from public controversies. “The most contested topics within Wikipedia are not things about the broader culture wars,” Wales, the site’s co-founder, wrote in an e-mail. “Of course there are difficult topics in the world, but most of the work isn’t about that. We have robust debates about even very obscure matters. That isn’t a problem. It’s what it means to be Wikipedia.”

Those debates unfold under unusual conditions. Anonymity shapes that process. Because editors generally operate under pseudonyms, an argument succeeds not on the strength of someone’s credentials or professional status but on its evidence and adherence to Wikipedia’s policies. Authority emerges slowly, through repeated participation. “You can’t force it,” Knapp said. “You earn it by being there.”

Pruitt is not known for winning disputes or shaping consensus through argument; he is known for his persistence. He fixes things quietly, avoids unnecessary fights, and doggedly returns to issues within an article that others have left behind. As Knapp said, “I think everyone agrees he’s a nice guy and does valuable things and is useful.” His scale of activity might suggest a certain level of influence over the site—he has altered millions of English-language Wikipedia articles—but the reality is more complicated. Wales wrote, “I just looked and this morning in the space of one minute he made 25 edits to remove a deprecated template parameter. He looks on pace to do hundreds of those in a day or two. That’s incredibly important, but hardly represents any sort of undue influence.”

That’s not to say that Pruitt doesn’t command an enormous amount of respect in the Wikipedia community. “We think of him as our No. 1 in amount of edits,” Rosie Stephenson-Goodknight, another super editor, told me. Stephenson-Goodknight, who is seventy-two years old, began editing Wikipedia in 2007, after looking up a defunct American publisher whose books she collected. “I thought there’d be a Wikipedia article,” she said. “And there wasn’t.”

Stephenson-Goodknight remembers double-checking the spelling of her search term. Nothing appeared. “It was that moment of clarity,” she said. “There is no Santa Claus. Mom and Dad are not infallible. Wikipedia does not have an article about everything.” So she decided to write one. “I started the article on Book League of America,” she said. “And it’s still there.”

Within a week or two, she wrote another. Then another. “I was hooked,” she explained. “It was like I was born to do this. There was just something about this researching and writing that just felt like it was me.” By 2014, she had become a super editor. Then, in December of that year, an academic paper quantified what editors like Stephenson-Goodknight had long suspected: just 15.53 per cent of English-language Wikipedia biographies were about women. A few months later, Stephenson-Goodknight launched a project called Women in Red, named for the red links that mark missing articles. The project focussed on creating Wikipedia articles about notable women. It has since produced hundreds of thousands of new entries.

In 2021, Stephenson-Goodknight was elected to the Board of Trustees of the Wikimedia Foundation, a position that she held through late 2024. Her tenure coincided with a fundamental shift in the role that Wikipedia plays on the internet. As the site entered its third decade, and artificial-intelligence algorithms grew hungry for data to learn on, Wikipedia articles were no longer just read; they were scraped, summarized, licensed, and folded into systems designed to answer questions elsewhere. In other words, Wikipedia had become infrastructure.

Pruitt was vaguely aware of the change before he fully grasped its implications. “Friends in tech would mention it,” he recalled. “They’d say, ‘You know Wikipedia is being used for this now.’ ” He did not follow developments in A.I. closely. “I don’t understand half of what Silicon Valley does,” he said. What he does understand well is reference works.

In October, 2025, when Elon Musk’s company xAI launched Grokipedia, an A.I.-generated encyclopedia that is often compared to Wikipedia, Pruitt approached it the way he would any new compendium. He searched for articles on subjects he knew well, such as nineteenth-century opera singers. “They weren’t there,” he said.

But what unsettled him was not what was missing from Grokipedia but what was slightly off. “Nothing was exactly wrong, but it was just less right than I would have made it,” Pruitt said. He described reading an entry that repurposed information from Wikipedia while subtly distorting it. In one instance, the entry summarized part of a person’s life in a way that struck him as careless, describing a seven-year period as “brief.” “I don’t think that’s brief,” he said.

The problem, as Pruitt saw it, was not the errors themselves—there are plenty of mistakes on Wikipedia—but rather where the responsibility for those errors lay. “Wikipedia can be fixed,” he said. Errors are corrected publicly, and editors can debate them. Responsibility is shared, and each edit can be traced. Grokipedia, on the other hand, and A.I.-generated information more broadly, obscured the information-gathering process. It generated text that sounded authoritative without revealing exactly how it arrived there. “It sounds right,” he said. “And that’s worse.”

Inside Wikipedia, the response was pragmatic. Editors clarified existing policy. The use of large language models to generate contributions or rewrite content became grounds for blocking users. The decision wasn’t simply ideological. “You can’t collaborate with something that doesn’t take responsibility,” Pruitt said. Responsibility, on Wikipedia, meant that every edit was logged, every disagreement archived, every change reversible.

Still, Wikipedia’s ban on A.I. doesn’t mean that A.I. is prevented from reusing content on Wikipedia. Multiple A.I. companies have acknowledged that they rely on Wikipedia as a source of training data and reference material, citing its scale and accessibility. The site’s content is openly available under a Creative Commons license, a feature long intended to encourage educational reuse but one that also made it especially legible to automated systems. Wikipedia became a foundational text for machine-learning models designed to summarize and answer questions using information drawn from other online sources. Analyses of A.I. outputs suggest that Wikipedia accounts for roughly a quarter of cited sources, according to data compiled by Semrush, a search-optimization company, with the Wikimedia Foundation, the nonprofit that hosts Wikipedia, announcing earlier this year that Amazon, Meta, Microsoft, Mistral AI, and Perplexity had joined its commercial service for large-scale access to Wikimedia data.

Wales, Wikipedia’s co-founder, doesn’t see this as a big change in the way that the site was already functioning. In his view, the relationship between Wikipedia and the broader information ecosystem has always been cyclical rather than hierarchical. “Wikipedia informs journalists who then inform billions of people,” he wrote. “Wikipedia informs academics who then inform billions of people. And the journalists and academics (and artists and poets and everyone else!) inform Wikipedia. We’ve always been a layer.”

For Wales, the central question is how A.I. is used. “The key for me is ‘human in the loop,’ ” he wrote. “If you use AI in your work, that can be good or bad—it’s up to you to do it well. We don’t believe that at the current time, AI is anywhere good enough to directly write for Wikipedia without very careful human oversight.”

Midway through Wikipedia’s twenty-fifth year, the site remains dependent on human volunteers: editors like Pruitt, who are willing to spend thousands of hours making meticulous changes, and who insist that they could stop at any time. “I could,” Pruitt told me—stop, that is. “If I wanted to.” The statement wasn’t defiant; it was matter-of-fact. ♦