In the age of ChatGPT, is Wikipedia in danger of becoming an invisible infrastructure scraped together by AI, rather than the destination Google searches have led us to for the past 20-odd years?
That’s one of the things we think about a lot. Wikipedia is licensed free of charge. This is our gift to the world and we are happy to be part of the public infrastructure. But we need support, so it’s important that people understand that the information (in the AI Overview) comes from Wikipedia. It’s not good if we become invisible. This is not necessarily the case, but we need to keep an eye on it. According to Pew Research, Wikipedia has a 3% chance of appearing on the first page of search results. At the same time, we found that AI summaries linked to Wikipedia 6% of the time. In that context, we are still in the spotlight. People still look at Wikipedia. However, that doesn’t necessarily mean they’ll click through. Our model is not based on page views or clicks, but on people remembering us when it comes time to consider making a small donation.
Are AI-driven changes starting to impact your efforts to get people to donate?
No, it’s not. Traffic from humans decreased by 8%. However, this does not mean that the overall traffic will decrease by 8%. That’s because AI bots are attacking the site. The slowdown in foot traffic has not led to a decline in donations. People love Wikipedia. But obviously we have to keep an eye on things.
Did you find a solution to make it more visible within the AI overview?
Our content is freely licensed, but attribution is required.
Google has always been positive about that. Some of the other players may not be as much. They need to be more careful about how their attributions work. It’s an ethical issue. It is important to quote. Where did you get this information? We expect even teenagers writing papers in school to footnote and cite sources.
There have been several recent cases where we have lost legal battles. In India, there was a ruling earlier this year that required volunteers to be identified. Then there are the unsuccessful legal challenges to the UK’s Online Safety Act.
we will never give up. We were blocked in Türkiye for three years, fought all the way to the Supreme Court, and won. In the UK, we lost in court because we pre-emptively challenged this law. The regulators hadn’t done anything to us yet. In fact, there was language in that ruling that suggested the government needed to pay attention to Wikipedia. We’re not trying to fight anyone. But the idea of identifying volunteers has no method. If they try to block us, good luck. Any country that does such a thing would be like China, and its people would be furious.
It may sound very aggressive, but it’s not. This is Wikipedia, and everyone loves Wikipedia. The reason we are getting into these troubles is because the legal definitions are too vague. For example, are we social media? Well, not really, but yes, we are social and we are media. But we function in a completely different way. Here’s what we’re proposing to legislators: Think before you pass internet regulations. Will this hurt Wikipedia? If so, you might want to think again.
What is your reaction to Grokipedia?
I didn’t have time to follow it up because I was busy promoting my book. Of course I saw it, but I didn’t have time to look into it deeply. But news reports seem to suggest much the same as I expected. Lots of hallucinations and lots of repetitive or low quality information. This is what you would expect from a large-scale language model (LLM). The question is whether people trust it. Elon (Musk) claims it will be more neutral than Wikipedia. As many have pointed out, Grokipedia seems to agree with Elon Musk on quite a few things, but that may not be the best definition of neutrality.
You’ve been under a lot of pressure from governments trying to influence the narrative. How do you and Wikipedia navigate it?
We have always been very principled not to submit to censorship anywhere in the world. In fact, this is helpful because the government knows it can’t force Wikipedia to do anything. We will not respond. If that happens, you’ll have problems with your hands.
From a product perspective, it’s about thinking about quality sources and encouraging and fostering thoughtful dialogue. If you’re looking up citations for a Wikipedia article, you’ll probably prefer the New England Journal of Medicine over a tabloid for things like medical information. It’s obvious if you take a step back and think about it.
But in an age where people tend to follow whatever comes into their feed, it’s actually really important to say, “No, wait a minute.” Think about where you want to get your information. We are very outdated in how we make such decisions. We consider things like whether the source location has an editorial process. If something is wrong, will they fix it? All the classic things that everyone thinks are part of good old journalism, we try to take that and live by it.
One of the areas where it’s becoming very difficult, and this is a huge problem all over the world, is the real breakdown of local journalism. You cannot write the history of a small town in India without a local newspaper to write the first draft. I wish I could join a massive campaign to find a way to solve it. We should all be concerned about that.
You were planning to come to India a few weeks ago, but I heard you haven’t gotten your visa yet.
I mean no offense, but I just wrote that it was the fault of India’s bureaucracy. When I applied for a business visa, I was told, “No, you need a conference visa.” And those at the meeting said they didn’t have the authority to issue the visa documents. They asked if I could apply for a tourist visa. When I did that, the bureaucrats said, “No, you’ve already applied. You already know it’s a conference visa.” I don’t think it was political. I love India and coming to India is one of my favorite things. I doubt (whether it’s political or not). I’ll be attending the Jaipur Literature Festival, so if you can’t make it, ask me in a few months.
Wikipedia is very popular in India. But what does Wikipedia need to change structurally and culturally to better serve the Indian community?
We need to focus more resources than ever before. We are moving in this direction. We have local chapters around the world, which receive funding from the Wikimedia Foundation and also conduct their own fundraising efforts. And my own opinion is that we need to continue to increase funding for small language chapters and really support them in building their communities and organizing events and activities. I think we should double that.
A key argument in your new book is that trust needs to be built into the system from the beginning. But what about existing, entrenched tools like social media? What do you think the big platforms, metas, and Xs around the world need to fix?
For these companies, this is a very complex issue. Social media’s ad-only business model means that algorithms try to show users as many ads as possible, promoting certain types of content that are addictive and outrageous. Click on a short video to see another video, and then another. After 30 minutes, you’ll think, “Wow, I could have done something useful with that time.” In some cases, it can actually be unhealthy and toxic. I quit Twitter (X) from my phone because I don’t think it will improve my life or make me happy. I still have the account because I need it for work. But I don’t go there for fun or for enlightenment or enrichment. These points need to be fixed, and more attention needs to be paid to the quality of the algorithm’s content. Show me some things that might not be as addictive, but might be a little more in line with what I’m really trying to accomplish in life. In my case, I want to learn something.
