Jimmy Wales doesn't love the word power. When Paxton Gray asked him where Wikipedia's outsized influence on the internet comes from, Wales pushed back on the framing. "Normally power is towards some end," he said, "and that's not really how I think about it at all. We're just a bunch of nerds writing an encyclopedia."
That deflection is actually the point. Wales, who founded Wikipedia in 2001 and now runs the Wikimedia Foundation, sat down with Paxton to talk through his book, The Seven Rules of Trust, and what 25 years of building one of the most visited websites on earth has taught him about trust, community, and the coming AI storm.
Wales is careful about what Wikipedia's influence rests on. It's not traffic or reach or brand recognition. It's trust, and not a naive kind.
"That doesn't mean trust that Wikipedia is perfect or that everything is a hundred percent correct," he said. "But trust that it's honest, that we're trying to get it right, trying to do the right thing."
That distinction matters to him. Wikipedia is credible because it's transparent: sources are cited, editors are visible, and the talk pages behind every article show the arguments that shaped it. People accept Wikipedia's information not because they're told to, but because they can verify the process themselves.
The nonprofit model reinforces that. Wales said there's something psychologically significant about being asked to donate rather than shown an ad. When users understand that Wikipedia is reader-supported and ad-free, it shifts how they read the content. "They're here to give me information. They're not here to promote whatever advertisers think."
He also noted that donors tend to become advocates. Once someone chips in, they want to tell others about it. Loyalty compounds. Wikipedia's email campaigns to past donors, timed around donation anniversaries, are one of its most reliable revenue sources, and the amount a user donates matters less than the relationship it builds.
For a site that draws hundreds of millions of readers, Wikipedia's KPIs are surprisingly mission-driven. Wales said the team watches growth in developing-world languages closely, treats quality and neutrality as primary measures, and monitors readership, though not as a fixation.
"We're a community of nerds who like writing an encyclopedia as a hobby. So that's what we're here for."
On the editor side, Wales is deliberate. He acknowledged that growing the editor base through cheap tricks, like a bigger edit button or pop-ups prompting users to contribute, would be easy and counterproductive. What Wikipedia needs isn't volume. It's the right people: editors with a commitment to neutrality and to treating others well.
"We're really looking for the right kind of people."
Most editors find Wikipedia the same way readers do. They arrive through a search engine, start reading, and eventually learn the site is editable. Some fix a typo. Some dive deep into a niche they know well. Others end up becoming admins or working on editorial policy. The funnel is organic, but the culture that catches newcomers is intentional.
The rise of AI-generated search summaries has put a dent in Wikipedia's traffic. Wales said the most recent numbers showed roughly an 8% drop in human visitors, driven largely by people getting quick-answer queries resolved directly in Google's AI Overview without ever clicking through.
He's not panicking. "It's important, it's relevant. It's not a disaster for us."
His reasoning: the lost traffic is mostly low-depth. If someone searches "how old is Tom Cruise" and gets the answer without visiting Wikipedia, that's a query Wikipedia wasn't really doing much with anyway. The deeper reading, the rabbit holes, the clicks from one article to five others, that behavior is still happening.
The contrast he drew was Stack Overflow, a programmer Q&A site that used to be the go-to destination whenever a developer hit an error message. Now AI tools answer those questions directly, and well. Wales said Stack Overflow's traffic has dropped around 90%. "That's a for-profit business. That's pretty devastating for them."
Wikipedia's structure insulates it from the same dynamic. Revenue doesn't come from ads tied to page views. It comes from donations, many of which arrive through email campaigns. As long as people know what Wikipedia is and care about it, the organization can sustain itself even as traffic patterns shift.
Wales made a point that most origin stories leave out: Wikipedia became what it is in part because it ran out of money at exactly the right moment.
In the early 2000s, the site was growing fast. Without the dot-com crash, the natural move would have been to raise venture capital and hire staff. Community managers, administrators, editors-in-chief. A top-down structure, like most social platforms.
"There was no choice but to innovate around how the community worked," Wales said. Without money to hire moderators, the only option was to build a community that governed itself. Admins elected by peers. Appeals processes. Cultural norms that calcified into genuine institutions.
"I don't recommend intentional deprivation as a means of innovation, but it does kind of work."
Had Wales raised $10 million in 2001, he said Wikipedia would likely look like YouTube today: a platform dependent on paid moderators, now wondering whether to replace them all with AI.
Pew Research estimated that AI bots and non-human agents generated more than 88 billion page views on Wikipedia in 2025. Wales said the reality is both flattering and complicated.
The flattering part: Wikipedia is foundational training data for most large language models. Its sourced, neutral, fact-based content makes AI systems better. Wales put it plainly: "You wouldn't really want to use an AI that was only trained on Twitter. It would be very stupid and very angry."
The complicated part: those bots cost money. Server load is real. And while everything on Wikipedia is freely licensed, with no copyright issues, attribution is another matter. Wales said AI companies are getting better at citing sources but still have room to improve.
The bigger issue is fairness. Wikipedia runs on $10 donations from readers who want to keep the internet good. Those donors didn't sign up to subsidize billion-dollar AI companies training models for free. So Wikimedia has built an enterprise API product and begun blocking crawlers from companies that won't pay for structured access. "You should pay your fair share."
Wales is also watching what agentic AI means for traffic. When someone runs a research agent, it might visit Wikipedia dozens of times in a single session on behalf of a single user. That user isn't hammering the server commercially, but the aggregate volume could add up fast in ways that are still hard to model.
Paxton Gray noted that one of the most common requests our agency gets from clients is help editing their Wikipedia page. Wales's reaction: "That's the right answer" to turning it down.
The right path, Wales said, is the talk page. If information is outdated or missing a key development, leave a note with sources. If a lawsuit was resolved favorably and the article still reads as if it's pending, flag it with a link to the outcome. That's legitimate, useful, and consistent with how Wikipedia is supposed to work.
What doesn't work is trying to make the article more favorable. The community watches for conflict-of-interest edits. Getting caught doesn't just fail; it actively makes things worse. Wikipedia editors become skeptical of any subsequent correction, even a genuine one. "If they try, the community gets very upset and then you get a little message saying someone with a conflict of interest has been editing this, and you actually make it harder to correct than a genuine error."
The better approach is simple: act like a proper person. Come in with a factual error, a source, and a neutral tone. That works.
Wikipedia's community culture has stayed remarkably stable over 25 years. Wales said the broad principles, neutrality above all, have never really shifted. But the detailed rules have evolved constantly, shaped by new problems no one anticipated at the start.
Early Wikipedia didn't need citations for common knowledge. As the site grew, and as more people tried to use it for PR or spam, sourcing standards tightened. Questions about which companies merit a Wikipedia article, which blogs count as valid references, and how to handle topics where mainstream journalism has dried up, these are live debates the community works through case by case.
Wales described a recent argument over a programming language with some users but limited third-party coverage. The challenge: most technical knowledge now lives on the web in first-party sources, not in books or magazines. The old sourcing rules may not fit that world. "Maybe we need to think about what we're using as sources in that area."
He sees this as healthy. The principle stays fixed. The application adapts.
Wales pushed back on the idea that AI is only accelerating the collapse of local journalism. He thinks the picture is more complex.
Local news has been dying for decades, well before large language models arrived. His hometown paper in Huntsville, Alabama, now publishes three times a week and is edited out of Birmingham, 100 miles away. The number of local journalists is a fraction of what it was when he was a paper boy as a teenager.
His mother read on the local paper: "Why would I subscribe? The newspaper's terrible." A chicken-and-egg collapse. The product got worse, the readers left, the revenue dropped, the product got worse.
But Wales sees a possible role for AI on the other side. He gave the example of building permits: a public database that used to require a lot of manual effort to monitor. An AI assistant could surface the ten largest new projects each week, flag which ones drew community objections, and hand the journalist a story lead. The journalist still does the work. The AI just removes the hours of sifting.
"If we can reduce the cost of doing some kinds of journalism, maybe they can start to build products that people actually want to support."
Wales turned 60 this year and said predicting 100 years ahead is nearly impossible. He's more focused on the next 25, and even that's hard.
His core argument for Wikipedia's staying power is that its editors aren't there for the audience. They're there because they like doing it. He contrasted this with Instagram influencers who leave the moment the audience does. Wikipedians don't have that problem. They're writing because they're nerds who enjoy it, regardless of traffic trends or AI summaries eating into page views.
"Wikipedia will still be here," he said. "I just don't know what it'll look like."
Paxton Gray's final question: has common knowledge grown or just changed?
Wales said both. The human brain hasn't changed. What fills it has.
His father was a shade-tree mechanic who rebuilt carburetors. That knowledge was common once. Now it's rare. People used to memorize phone numbers. Now no one does, and Wales is clear: that's not stupidity, it's efficiency. "You don't need to know people's phone numbers anymore. You just look it up. That's just silly."
The harder question is whether we can process new information without a strong base of existing knowledge. Wales thinks we still need that base badly. He used a recent news story about the possibility of bombing Iran's desalination plants. To evaluate the idea, he needed to know where Iran gets its water, how other Gulf countries compare, and what the strategic implications of escalation would be. None of that is common knowledge, but all of it is context without which the headline is noise.
"You can't even understand it unless you've got a pretty broad base of knowledge in your brain as context."
The point he kept coming back to: access to facts has never been easier, but the ability to use those facts still depends on a foundation built over years.
Get Jimmy’s new book The Seven Rules of Trust: A Blueprint for Building Things That Last: https://sites.prh.com/the-seven-rules-of-trust
Follow Jimmy on Twitter: https://x.com/jimmy_wales
Connect with Paxton on LinkedIn: https://www.linkedin.com/in/paxtongray/
Looking for an agency that'll be worth the investment? 97th Floor creates custom, audience-first campaigns that drive pipeline and conversions. Get started here: https://97thfloor.com/lets-talk/.
Jimmy Wales is an American-British Internet entrepreneur best known for founding Wikipedia.org in 2001, the Wikimedia Foundation in 2003, and Fandom in 2004. Today, Wikipedia and its sister projects are among the top-five most visited sites on the web.
He holds finance degrees from Auburn University and the University of Alabama and was appointed a fellow of the Berkman Center for Internet & Society at Harvard Law School in 2005. His debut book, "The Seven Rules of Trust," was published by Bloomsbury and Crown Currency in October 2025. Wales has been recognized with the Time 100 Award, the World Economic Forum's "Young Global Leaders" designation, and the UNESCO Niels Bohr Medal.
