Home → Other → Interviews
In an age of AI hallucinations and collapsing trust, Jimmy Wales argues that Wikipedia’s human transparency may be exactly what makes it worth trusting.
For much of its 25-year life, Wikipedia has lived with a big question mark beside its name.
Teachers warned students to stay away from it, and many experts doubted it. The very idea sounded absurd: an encyclopedia that anyone could edit? Yeah, right, like that’s gonna work.
But the internet changed around Wikipedia.
Fast forward to today. We live with AI hallucinations, search results shaped by opaque systems, and algorithms that often reward anger over facts. In this world, Wikipedia’s old weakness has started to look like a strange kind of strength. It feels unfinished and argumentative, but deeply human. Unlike much of the modern internet, it lets readers see the argument behind the answer.
Jimmy Wales, Wikipedia’s founder, sees that transparency as central to the site’s credibility. I met Wales at the 2026 Cheltenham Science Festival in the UK and asked him whether Wikipedia’s imperfect, volunteer-driven nature has become part of why people trust it — especially now that AI systems can deliver answers that are polished, persuasive, and wrong.
“I think that does give rise to trust,” Wales told ZME Science. “The fact that you can click on the talk page and see what the Wikipedians are arguing about, that’s useful.” Whereas AI chatbots provide opaque information, Wikipedia’s system gives everyone a chance to see how the information was added.
Where AI chatbots often give answers without showing their workings, Wikipedia leaves a detailed trail. You can go on any page and see the sources, the edit history, the disputes. You can see where people disagreed, and sometimes, why one version survived.
“We’re very happy to tell you all the things that are wrong with Wikipedia and all the criticism we’ve gotten cause that’s just part of history. It’s part of the process,” he told me.
But this is not only a story about Wikipedia. It is a story about trust in an age when the systems shaping our world often feel distant, automated, and hard to inspect.
Before Wikipedia, Wales was trying to build a free online encyclopedia the old-fashioned way. It was carefully crafted and top-down, relying on expert knowledge. That project, called Nupedia, never really worked.
Then came the wiki model: a simple tool that allowed people — anyone really — to edit pages directly. Suddenly, the project opened up. Wales later recalled that moving away from “a previous model which was very old-fashioned, very top down” created “this burst of activity” from volunteers. The team got more work done in two weeks, he said, than it had in almost two years.
Stay ahead with ZME Science and subscribe.
Please check your inbox and confirm your subscription.
Things only grew from there. What began as a practical shortcut became Wikipedia’s defining idea. Knowledge would not be handed down by a small group of authorities. It would be built in public by anyone willing to help, using a transparent process.
At first, that made Wikipedia look unreliable but, in today’s context, it’s increasingly starting to look radically human. In a world where many people distrust institutions, Wikipedia offers a different kind of authority. Not the authority of certainty, but the authority of an inspectable process and a sort of wisdom of the masses.
Wales, who has recently written a book on trust, sees that lesson reaching far beyond the encyclopedia.
“We’ve seen this enormous decline in trust across society. Decline of trust in journalism, in politics, in business, in each other. It’s really become a crisis of trust. And at the same time, in 25 years, Wikipedia has gone from being kind of a joke that people didn’t believe in to one of the few things people trust, even though it’s imperfect and it’s always got problems and so on,” Wales said in our ZME Science interview.
Wikipedia is one of the rare internet institutions that seems to have become more trusted with age. It was widely mocked and criticized in its early years. But more recently, it’s been increasingly praised, including by researchers who study its accuracy, citations, and public role.
Part of that shift comes from refined and incremental improvement of the knowledge base, Wales says.
“There was a time where you could have been the first person to write ‘Paris is a city in France’ and hit save and that that was the article about Paris. So that wasn’t very good and it might not be reliable. But we’ve just been hard at work for 25 years trying to make Wikipedia better. “
But another part comes from the process itself.
Traditional institutions often try to inspire trust by projecting authority. They have experts, mastheads, offices, titles, and so on. Wikipedia has very little of that.
For starters, its pages are never truly finished. They’re working documents, constantly edited and sometimes fiercely debated. That was once the case against it. How could anyone trust an encyclopedia that anyone could edit facts in or out of?
But on today’s internet, Wikipedia’s unfinished quality can feel oddly refreshing. A reader can see not only what a page says, but how it got there: the sources behind a claim, the edits that survived, and the disputes around it.
That matters even more in the age of generative AI. AI usually sounds very confident even when it’s wrong. It produces clean, fluent explanations with confidence but without making clear where the information came from or how it weighed competing claims. It can be useful, but it is often a black box.
For Wales, the crucial point is not that humans are always wiser than machines. It is that humans can challenge one another in public, and that challenge becomes part of the record.
“If a chatbot gets something wrong, well, it probably didn’t mean to. It probably just hallucinated. Or it was just said what it was trained on… But that human element of really challenging things, really chewing on ideas… as long as we take that healthy attitude, we’ll always be doing great,” Wales says.
Of course the irony is that virtually all these chatbots have also been trained on Wikipedia. Yet surprisingly, Wales doesn’t seem too fazed by that.
“It’s fine. It’s all part of the grand march of technology and we definitely think it’s a good thing if AI chatbots are not only trained on social media or some nonsense. Wikipedia is a very valuable resource for that.”
Wales sees Wikipedia as more than a successful website. He sees it as a model for public life.
Wikipedia works because people can disagree inside a shared framework. They agree on what sources of information are acceptable, that corrections are possible (and necessary), and that neutrality is an aspiration, even if it is never perfectly achieved. Plus, no one person gets to declare reality by force.
The rest of the internet (and the rest of our society) isn’t really like that.
Social media plays a big part in this problem, Wales points out.
“Certainly we can point to the toxic nature of a lot of social media. When you go on social media you tend to get very negative feedback and negative views because that keeps people engaged. They want to argue with someone on Twitter and it’s a waste of time but there you are.” He also singles
The crisis goes beyond platforms. It also affects journalism, science, politics, and public debate.
“There are definitely politicians who are deliberately and very loudly trying to undermine journalists and journalism. That’s obviously a problem because we need journalists and we need journalism.” It’s one thing to legitimately criticize a newspaper for having a bias, for instance, and it’s another to say a newspaper is “fake news and complete trash,” Wales points out. “That’s ridiculous and not helpful at all.”
That idea, that we need a shared understanding of reality, is exactly why Wikipedia matters in this story. The site does not end arguments. It gives arguments somewhere to go; it gives a shared based of reality, and off a clear trail showing where the information comes from.
If we reach a point where we don’t agree on a shared basis of facts, then any debate becomes pointless, Wales concludes.
If Wikipedia offers one model for trust, it’s not an easy one to replicate.
You can’t apply its model to social media, for instance, where speed and outrage are often rewarded more than patience and facts. Wales doesn’t pretend there is a simple fix. I asked him whether other platforms could be built around trust rather than engagement.
“I think so, but I think it’s a hard problem. I will concede. You know it’s easy to criticize. It’s much harder to offer a more positive way forward,” he said.
Still, he does not see the design of online platforms as inevitable. Some companies, he argues, understand that pushing users toward darker, angrier places isn’t just bad for society. In the long run, it’s also bad for business.
“I feel like YouTube has been aware for a long time that if they aren’t very careful, their algorithm will start to promote really bad and negative things and send people down dark places,” Wales said. “And they do work to combat that because they have an understanding that actually for our long-term business model, you know, people thinking we’re destroying society isn’t actually very good for us.”
He’s not saying YouTube, or any platform, has solved the problem, but he thinks they’re at least trying to steer away from the worst incentives. Others, in his view, are not.
“Some just don’t care at all,” Wales said. “They really just are very happy to put forward [whatever]. Maybe they are ideological themselves and really want to promote certain views and they’re happy to use their platform to do it,” he says, before calling out Elon Musk as an example. “I just think that’s not probably what we really need right now.”
The contrast with Wikipedia is striking. Wikipedia is far from perfect, and Wales is the first to say so. But its incentives are different. It doesn’t need users to stay angry and it doesn’t reward the loudest. Its basic promise is smaller and more transparent: here is a claim, here are the sources, here is where people argued about it, and here is how it can be changed.
Wales has another piece of advice, and it is simpler than any platform redesign or algorithmic fix. In his new book, The Seven Rules of Trust, he says the first rule is to “make it personal.” By that, he means remembering that trust is built between people before it becomes a property of institutions.
“Make it personal is rule number one in the book and it’s really about thinking about the other person who’s involved and how they’re feeling? What are they doing? Because you’re much more likely to build trust if you can understand the people around you,” Wales said.
That may not be enough to repair a fractured public sphere. But it offers an approach. Trust is not rebuilt by demanding that people believe. It is rebuilt by giving them a process they can inspect, challenge and improve.
This article is based on an interview carried out at the Cheltenham Science Festival. The interview and its transcript follow below.
Interviewer [Andrei Mihai, ZME Sciencec]: Thank you so much for taking the time. Could you please introduce yourself for our audience?
Jimmy Wales: Yeah. I’m Jimmy Wales, founder of Wikipedia and the author of The Seven Rules of Trust, which is what I’m here in Cheltenham to talk about today.
AM: So, you wrote a book about trust. Why trust and why now?
Jimmy Wales: Well, we’ve seen this enormous decline in trust across society. Decline of trust in journalism, in politics, in business, in each other. It’s really become a crisis of trust. And at the same time, in the same sort of 25 years (Wikipedia is 25 this year) Wikipedia has gone from being kind of a joke that people didn’t believe in to one of the few things people trust, even though it’s imperfect and even though it’s always got problems and so on. So I thought maybe some of the lessons we learned in terms of building trust, not only sort of trust with the readers, but also trust with each other, within the community, might be helpful. So I thought I would write a book.
AM: So I want to quote from a Wikipedia page called the reliability of Wikipedia and it says surveys have had mixed results and studies but Wikipedia’s reliability has been frequently criticized in the early 2000s but has improved. It’s in the 2010s and early 2020s it has been generally been praised including in peer-reviewed studies. Why do you think that is? Why do you think people trust Wikipedia more than they did 20 years ago or 10 years ago?
Jimmy Wales: Well, it’s gotten better. That’s one thing. I mean there was a there was a time when you could have been the first person to write “Paris is a city in France” and hit save and that was the article about Paris. So that wasn’t very good and it might not be reliable. But we’ve just been hard at work for 25 years trying to make Wikipedia better. We’ve got a great community of people who care about reliable sources and trying to get it right. And so as it’s gotten better, that’s really important. I love an article like that because it shows the transparency of Wikipedia. We’re very happy to tell you all the things that are wrong with Wikipedia and all the criticism we’ve gotten cause that’s just part of history. It’s part of the process.
AM: Yeah. So, as you said, Wikipedia is 25, right? The world has changed a lot and the internet has changed tremendously in those years. Wikipedia was competing with expert written encyclopedias in the beginning and search engines and so on and now it’s sort of competing with LLM chatbots, right? Who scrape the content and then send users the same content or variations of the same content. Do you think that’s okay?
Jimmy Wales: Yeah, I mean it’s fine. It’s all part of the grand march of technology and we definitely think it’s a good thing if AI chatbots are not only trained on social media or some nonsense. You know, Wikipedia is a very valuable resource for that. Obviously, we’ve got you know some issues. So certainly people who want to contribute to Wikipedia should be very cautious and and probably should not try to use an LLM to try to write something for Wikipedia because they hallucinate. They get things wrong. They also have a very annoying style. Then you know, obviously, the traffic question is foremost. We didn’t see any impact really from the first big boom of ChatGPT because people were using it for all kinds of different things. I think we’re seeing some impact now from the AI summaries at the top of Google, but it seems to be mostly the short form, the quick clicks. You know, people just coming to ask Google a question, Google knows, so they just don’t need to come to Wikipedia for that. But we still see a lot of people coming for the long reads and for really exploring a topic and so on, going deep and going to the sources and all of that. The traditional use of Wikipedia is not one that an LLM replaces.
AM: But are you concerned? Do you think it’s sort of an existential threat LLMs or is it just something that it’s part of the process like you say?
Jimmy Wales: I think it’s just part of the process. I mean certainly if traffic were to fall by 90% that could be an issue for fundraising but I don’t think it would be an issue for the community because we’re a bunch of nerds who write an encyclopedia as a hobby. So we’re going to carry on regardless. We have to be thoughtful about what is our place in the internet? What is our place in the world? I mean there was a time when the rise of mobile presented some new interesting challenges for us because it’s very hard to edit Wikipedia on a mobile phone. Always will be even if we make it as good as possible. It’s you know tiny screen and all that. But reading in the early days was maybe not so good and that’s gotten to be really good. People read Wikipedia on their phones all the time. So these changes happen and we have to think about them, like how can we use them, what what is the right way. I’m very interested in thinking about are there ways that large language models can help us make Wikipedia better not by writing articles but by you know scanning through things making suggestions noting you know discrepancies between different things that could be useful we’re experimenting with some things like that.
AM: So you’re experimenting right now, but is there any AI feature actually implemented in Wikipedia editing in some sort?
Jimmy Wales: No not really. Not much. I mean we have a little bit machine learning to help identify vandalism and things like that. But in terms of large language models, not really. I mean the community is using different tools completely independently, they’re exploring they’re they’re learning things. So I’m sure some of them are using large language models in some interesting ways. But it’s not yet you know formally part of the process.
AM: You mentioned that Wikipedia has gotten better. Is there maybe also a sort of emotional response that people have? Because Wikipedia it’s imperfect as you say. It’s constantly building. It’s very human in a way, right? It’s just people arguing with each other. Whereas AI is LLMs are the opposite where they’re very convincing, but they’re also very opaque. There’s no transparency. Do you think people sort of appreciate that and trust that?
Jimmy Wales: Yeah, I think they do. I think it does give rise to trust. The fact that you can click on the talk page and see what the Wikipedians are arguing about, like that, is useful, you know. If a chatbot gets something wrong, well, it probably didn’t mean to. It probably just hallucinated. Or it was just said what it was trained on and so forth. But that human element of really challenging things, really chewing on ideas, really being open to someone coming and say, “Hey, I think you got this wrong.” And then you say, “Oh, hold on. Did we get this wrong? Yeah, let’s have a let’s have a conversation about that.” As long as we take that healthy attitude, we’ll always be doing great.
AM: Zooming out mentioning one of the things you said that there’s a crisis of trust worldwide. Do you think there’s an overarching theme that’s linking lack of trust or declining trust in politics and in journalism? Do you think there’s something in our society that’s causing that? Is it sort of I don’t know people just being more pessimistic or something else?
Jimmy Wales: I think it’s a few things. So certainly we can point to the toxic nature of a lot of social media, where when you go on social media, you tend to get very, very negative feedback and negative views because that keeps people engaged. They want to argue with someone on Twitter, and it’s a waste of time, but there you are.
Other things, there are definitely politicians who are deliberately and very loudly trying to undermine journalists and journalism. That’s obviously a problem because we need journalists. We need journalism and crossing the line from a legitimate critique or concern, like if somebody says the New York Times does have a liberal bias, it’s a bit left-wing… okay, that’s great, that’s a conversation. If you say “the New York Times fake news is complete trash”, that’s ridiculous, that’s not even helpful at all, and you’re undermining the very thing that we need as a society, which is a shared understanding of reality, a shared truth. In the book, I interviewed Christian Amenor, a very well-known and respected journalist, who says (I’m going to paraphrase what she says), “Look, we have to have a shared basis of facts, and then what you do about the facts, of course, you can debate about that”. But that shared basis of facts, if you don’t even have that, then the rest of the debate is absolutely pointless because you don’t even say you’re not even looking at the same reality, so this is very important.
AM: Yeah, and social media platforms are driven by engagement, and they always try to milk so much engagement, whereas you said Wikipedia is built around trust in a way, or it functions around trust. Do you think there’s a world in which social media platforms could be also built around trust? Could they be profitable and you know contribute to a healthier society?
Jimmy Wales: I mean I think so but I think it’s a hard problem. I will concede. It’s easy to criticize. It’s much harder to offer a more positive way forward. I do think they can do a better job and I certainly would say of all the various platforms some are better than others. I feel like YouTube has been aware for a long time that if they aren’t very careful, their algorithm will start to promote really bad and negative things and send people down dark places. And they do work to combat that because they have an understanding that, actually, for our long-term business model, you know, people thinking we’re destroying society isn’t actually very good for us. Do they always get it right? Of course not. Like it’s a big company. They’ve got a lot going on. But I do think they’re doing better than some of the others and some like they just don’t care at all. Like they really just are very happy to put forward [whatever]. Maybe they are ideological themselves (Elon) and really want to promote certain views and they’re happy to use their platform to do it. And I just think that’s not probably what we really need right now.
AM: Yeah. Going back to Wikipedia, you know, I think everyone’s a Wikipedia user more or less; and every now and then you see these messages like we’re fundraising and so on. It always seems like Wikipedia is vulnerable, like it’s always on the edge. Is that the case?
Jimmy Wales: We have to take fundraising seriously. You know, we’re doing fine. We are very careful financially as an organization. Despite being so big and a global brand, we keep a pretty small budget. You know, we try to manage things. We try to build our reserves and so on. Because we don’t know what technology is coming and all of that. So, you know, we don’t want to scare people, right? We think, you know, donate because you want to keep us healthy, not because we’re about to go out of business.
Interviewer: Okay. So Wikipedia also seems very unlikely. When you tell someone the story about how it started, it seems, at least to me, like an unlikely success story. Did you think it was going to succeed in the beginning? Was it something you were confident in?
Jimmy Wales: The thing is I always say I’m a pathological optimist, so I always think everything’s going to be fine. It’s just my nature. So, I was like, “Yeah, this sounds fun. This sounds great. Let’s do it.” So yeah I was always very optimistic and even in the very early days you know I saw when we switched to the wiki model from a previous model which was very old-fashioned very top down and suddenly this burst of activity came from the community and we got more work done in you know in two weeks than we had in almost two years. I was like this is cool, this is actually really interesting, we’re having this very open model where people can come and get started. We’ve unlocked people’s excitement and energy and this is fantastic. So yeah, even in the very early days when various problems arose and things like that, I felt fundamentally this is going to be okay.
AM: Yeah and I think that’s that’s pretty much or if you have any 10-second takeaway from your book if there’s one idea you want your people to take from your book.
Jimmy Wales: Make it personal is rule number one in the book and it’s really about thinking about the other person who’s involved and how are they feeling? What are they doing? Because you’re much more likely to build trust if you can understand the people around you.
Interviewer: Brilliant. Thank you!

Dr. Andrei Mihai is a geophysicist and founder of ZME Science. He has a Ph.D. in geophysics and archaeology and has completed courses from prestigious universities (with programs ranging from climate and astronomy to chemistry and geology). He is passionate about making research more accessible to everyone and communicating news and features to a broad audience.
© 2007-2025 ZME Science – Not exactly rocket science. All Rights Reserved.
© 2007-2025 ZME Science – Not exactly rocket science. All Rights Reserved.