How AI Is Breaking the Internet (and What to Do About It) | Matthew Prince | TED

The internet is an absolute miracle. And for the last 30 years, it has been driven by [a] very consistent and fundamental business model that has given us everything amazing about the internet and everything not so amazing as well. I run a company that provides a lot of the internet's infrastructure and as a result, means three things. One, we've had this unique view into what is going on online. Two, we care enormously about the success of the internet. And three, I can share with you some of how that underlying business model is changing and why it's important. Because as that business model changes, it's going to have a dramatic impact on what the internet is, but beyond that, on what content is created, on what journalists get compensated for, on how researchers are able to do their research. And so it's up to all of us as that business model of the internet changes, to make sure it changes for something we want and can learn from the mistakes that we had in the past.

So to review the past, let's look at what that business model was. And we all know it basically, which was: the internet was built on we took an idea, and we put it online, and then we drove traffic to that idea. And then from that traffic, we either sold things, subscriptions or ads. And it's been pretty amazing to watch how that has very much driven the growth of the web and the growth of the internet over time. And it survived a number of dramatic platform shifts. We all started browsing the internet on browsers, on our desktop or laptop. That transitioned to be increasingly social media, which was the interface through which we consumed content. And then we went to consuming it through our mobile devices. And yet, through all of this, the humble banner ad has basically financed it all. Traffic has been the currency of the internet, and that's what everyone has built their content to try to achieve. And if we want to look for who's the Superman, who's the superhero that has driven this for the last 27 years, that has been Google.

Google built not only the search engine to index how the internet works, but then they built the monetization system to actually ensure that content creators got paid for the traffic that they generated. The internet owes a ton to Google in the last 27 years. But the world is changing. AI is the topic du jour, and what I think a lot of people don't understand is that AI is yet another platform shift. Think about how much more of the way that you consume information already is looking at your AI chatbot, and how much more of that will be the case in the future. We can see it in a lot of things we do kind of normally, like, this is a Google search from the early days of Google. Not even that long ago, where I would type something in, like, when was the first TED conference? And then I would get back basically a treasure map where I would have to continue to do the work and click on the links, generating traffic to find what my answer was. The search wasn't over when the results returned. That was just the beginning, and you had to follow the treasure map.

That's what a search engine is. Today we have something very different. We have something that's closer to what I would call an answer engine, where you're getting the answer directly on the page. And what's amazing about that is that as a consumer, it's better for you. But what's terrifying about it is that it means traffic isn't going to the publishers' sites, and traffic is the currency of the internet. And so 18 months ago, publishers called us and said, “Are you seeing this, too?” And I have to confess, we weren't really paying a ton of attention, but we went into our logs, we went into our data, we pulled out that information and said, “How much harder is it to get traffic based on the same amount of content today than it was with the Google of old?” And the answer is, it's 5.2 times harder to get traffic for the same amount of content that was out there. But that's the good news. That's with Google. What is it with OpenAI? 750 times harder. What is it with Anthropic? 32,000 times harder. The answer engines are giving the answers and people aren't following the footnotes.

And this was 18 months ago. What’s happened since, has it gotten better or worse? Do you trust AI more today than you did 18 months ago? The answer is most people do, and as a result, it's actually gotten twice as hard to get traffic to content that's there. You're not following the footnotes anymore. You're actually trusting that AI systems are giving you the right answer, and that's destroying media. Across the board, we're seeing reported numbers ... of between a 90 and 30 percent decline in the traffic to their sites, and this is extremely problematic because I think journalists deserve to be able to eat. (Applause) And fundamentally, if the internet is built on traffic and this new business model isn't delivering traffic, then I'm not sure how media is going to survive unless we create a new business model. And the good news is the AI companies want journalism. They want people out there doing research. We all want journalism. We want people out there doing research. And so I think, I'm actually kind of optimistic that there might be a solution to this problem.

And here's the way I think about it. For the first time in human history, we have a mathematical model of all of human knowledge. It's not perfect, but that's essentially what all of the LLMs are. And I think of it like a giant block of Swiss cheese. There's a lot of cheese, that's what we know. But we also have a lot of holes in the cheese. Those are the things we don't know. And the thing that is incredible is that what all of you want, what I want, what the AI companies want is they want people out there filling in the holes in the cheese. They don’t want you writing stories that already exist -- they want you actually creating new knowledge. And the AI companies, I've talked to every one of them, they're all willing to actually pay for people who are creating new knowledge, creating new ideas, ideas worth spreading. That's what this new system would reward. Compare that to what the business model of media is for the last 10 years, which is not filling in the holes in the cheese, but instead, basically, lighting the cheese on fire, right? Incendiary cheese, right?

Yet another story about what happened at 1600 Pennsylvania Avenue, told in the most inflammatory way possible. I've sat with media executives who brag about the fact that they create content for as little as possible, and then write headlines that create the largest cortisol response so that you will click on them. That's not a world we should be celebrating. That's not a world that's making the world better. If, on the other hand, we actually celebrated who is making the cheese bigger, who's filling in that -- Imagine in the future that you actually had AI companies that were measuring who actually advanced human knowledge mathematically over the course of the last year the most, and we gave them an award. We could call it the Big Cheese award. (Laughter) I'm not kidding. Like, how amazing would that be? And you might be thinking, OK, Matthew, you're crazy. Like, I'm the person who thinks AI is just destroying the world, and technology always makes things worse. I'm going to share just two anecdotes that make me have some hope that this might actually be the case. The first starts 23 years ago. 23 years ago, the music industry had a huge problem: Napster, Kazaa -- everyone was stealing music.

And then Steve Jobs stepped on stage, almost 23 years ago to this day, and he said, “We’re going to do something different. We're going to charge 99 cents per song. We're going to actually say that you should be paying for the music that you consume.” Now that’s not the business model that won, but it was the start, and it turned into what is today Spotify and all the other systems that are out there, which is basically $10 a month, all-you-can-eat music, and it's pretty amazing. Last year, Spotify alone sent over $10 billion to music creators in revenue. And again, that's in addition to titles, TikToks, YouTube, Apple Music, everybody else, more money is going to music creation than any time in human history. So technology can be pie-expanding. In fact, I thought it was so important that I flew up to Stockholm to meet with Daniel Ek, the founder of Spotify, and he said, "You know, it's pretty amazing. We know when someone goes to Spotify and they type in Taylor Swift, "Shake It Off" that we can give them what they're looking for."

But every once in a while, someone comes in and types in something like, 'I want a song about how much fun it is to dance disco with your cat.' (Laughter) And nobody's written that. But they know that nobody's written that. And so they actually go back and take those queries and then publish them back to music creators. And this is the moment you're all going to think you're doing life wrong. But there's a guy who lives in Denmark who makes €40 million a year writing songs for unfulfilled Spotify queries. (Laughter) And I think that's beautiful. That's filling the holes in the cheese. Because when I look for a song, what I'm looking for is an emotion. And this person is helping me find a song to resonate that way. And I think that's a model for what could happen as well. Another story starts a little bit closer to home. So this is the Park Record newspaper. The first place I ever saw my name in print -- my hometown newspaper from Park City, Utah. It was dying. So three years ago, my wife and I bought it. We did not buy it because we thought it was going to be a great business.

We bought it because we thought local news was important. In fact, we had no financial KPIs -- the only KPI we focused on was how much could we increase voter turnout. And the answer was, we got it from the low 20 percent to over 60 percent in the last election in three years, which is amazing, right? (Applause) Local news matters. But here's the crazy part. Terrible business, right? Except it's not. We're likely going to make more in 2026 off licensing our content to AI companies than we do off digital advertising. Why? Because if you're traveling to Park City, Utah, and you want to know what's the great new restaurant, what's the cool band that's coming to town, we're the only ones that have it. So think about how amazing that would be. Think about how if in the future, the business model of the new AI-driven internet actually rewards the media we're all craving, local, unique, original content. That's what I'm super excited about. But there are problems. Challenges. It's kind of like in the superhero films, where the hero of the last movie becomes the villain of the next.

And so our good friends at Google ... Ugh. (Laughter) They are holding this back. They believe that they have a God-given right to all of the content online. They shouldn't have to pay for any of it. And they have a huge advantage. In fact, it's hard to believe this, but every other AI company was actually started to oppose Google, right? That’s what OpenAI started with. Anthropic was just a spinoff of OpenAI, xAI. All of this was all to say, we can't let Google win this entire game. How much of an advantage do they have? Well we looked at this again, and we found that every single page that OpenAI sees, Google sees three. Four in Microsoft’s case, five in Anthropic's case, 21 or 22 in xAI's case. An enormous advantage that Google has, and they want to hold on to that advantage. In fact, they're playing incredibly dirty. They're saying to content creators, "You can't stop us from training on your content or we will take you out of search." Horrible. Superman, how far you’ve fallen!

(Laughter) And so I was pretty depressed, I have to confess, thinking we were on the cusp of creating something amazing: a better internet. And yet, I have to just realize that no AI company's top priority is paying content creators. What's their top priority? Beating Google. Which gave me an idea. Kind of a wonderful idea. An idea which is going on right now, and it's working. Which is: don’t charge for all the content that’s out there, just charge the AI companies for the ability to catch Google. And I don't think that's the end game, but I think it might be the iTunes moment. It might be the point that shows us content is worth paying for. Everyone should be out there, and we can create a better business model for the internet. And so I'm hopeful that on this stage, not too long from now, I get to hand out one of the first Big Cheese awards, and I really appreciate it. Thank you.