What does the internet look like in a post-search world?

At midnight on July 1, near the top of One World Trade, Cloudflare CEO Matthew Prince pressed a red button to much fanfare from media execs in attendance.

The button didn’t look like much, according to The Rebooting media newsletter, but it was meant to symbolize a bold new chapter in how businesses that operate on the open web relate to AI. Having seen website traffic plummet from Google’s new AI search features and its chatbot competition, web infrastructure company Cloudflare was turning on a system that would, by default, block the crawlers that pull content to train AI, and implement a beta pay-per-crawl tool.

Prince has spoken at length about what might happen to the internet if web publishers don’t collectively stand up to AI companies—nothing good—and how the future of the internet could play out as AI chatbots supplant search engines.

We caught up with him to talk about what’s happened in the months since the button, how AI could actually reward local news, and why he seems relatively optimistic about all this.

This conversation has been edited for length and clarity.

So you had this event in July where you pushed this red button and you enacted this system. What have you seen since then?

We thought that this was still a very early step in what will be a long journey, but we’ve been very encouraged by the results, both in terms of the publisher community as well as early results within the AI community. In the publisher community, we had primarily focused, just mostly because of time, on US-based publishers. And so the thing that has been really striking since July 1 has been how this is a problem that is affecting content creators and publishers around the world…Everyone is seeing the same thing, which is [the shift] from search engines…to answer engines, which, I think everyone agrees, is a better user experience for 95% of users, 95% of the time.

So the world is going to shift to that. As that happens, everyone around the world is seeing a drop in traffic, and that’s hurting their businesses…We’ve now got a true global coalition of the publishing industry saying something has to change.

The big thing that’s happening on the AI side has been that the conversations around compensating content creators for content have gotten much more fruitful, and the rates that the AI companies are willing to pay have markedly ticked up. And I think that what the AI companies are seeing is that if you look at the state of the art models that are out today, they were trained basically through data that ended around January of 2025. The next generation of models that are coming out are now having to use data that is after July 1 of this year [what Cloudflare refers to as “Content Independence Day”]. And the early results that we’re hearing back from the AI companies is that they’re hitting a plateau in terms of their ability to further those models and make them better. If those models stall out, that will very clearly show the value that publisher content data is providing, and it will say that it’s not enough for you to pay for fleets of GPUs. It’s not enough for you to pay billions of dollars for AI researchers. You’ve also got to be paying for content if you want to have the best model that’s out there.

I was reading a recent interview where you were talking about three different outcomes for what AI could mean for the internet. One is that original content just kind of dies, which seems like obviously the doomsday one. But if AI companies do realize that content is important, you said there are two other possible outcomes: one where AI companies control the content production for training data, maybe start their own version of the Associated Press—

Can you imagine? Would OpenAI buy Gartner? It’s not crazy.

And then a third outcome where publishers are compensated. What do you think could tip the scales between those second and third outcomes, whether AI companies decide to just create their own content? What factors might determine whether they would want to do that?

The good news is that it’s in everybody’s interest for the third outcome, the new business model, to emerge. It is likely to be a cheaper and less risky outcome for the AI companies…I don’t think [AI companies] want to stand up their own fleet of journalists; that’s not the business that they think of themselves in, and so it’s pretty risky for them to do it.

From the perspective of users, it’s incredibly regressive to think that the future is a place where information gets incredibly siloed. If you subscribe to OpenAI, you’re only going to get what their own journalists get. I mean, inherently, that is going to be biased news sources, and over time, that will divide up the populace. There will be a conservative AI company, there will be a liberal AI company, there will be a Chinese AI company. And again, I think users—and in addition to that, regulators and policymakers—really don’t want us to go backward and re-silo information. The amazing thing the internet did is it democratized information. And so it seems against users’, regulators’, policymakers’ interests to do that.

I own a small local newspaper. The journalists that work for us would rather work on what they want to do. They would rather be closer to independent contractors who get to follow whatever stories they’re the most interested in, rather than having to do whatever Sam Altman orders them to do. And so I think that the journalists themselves would prefer a world in which they can get paid to be creative, as opposed to get paid to work for one of the big five AI companies.

So I think it’s in everybody’s interest to find a solution to this, which is why I’m optimistic that that solution will emerge, even if it feels like it’s such a radical change from where we are today. It is a radical change, but the user interface of the web is going to change radically because AI is such a radical change from search. And that’s going to drive this dramatic change and you would expect that it will drive an equally dramatic change to the business model.

You mentioned owning a local paper, the Park Record in Park City, Utah—and I don’t know how hands-on you are with day to day operations there—but are you thinking about how you can position it for this new AI future?

First of all, we didn’t buy the newspaper because we thought it was a great business. We bought the newspaper because we think that every community needs to have some local news source or you don’t have a community…There has to be some way that you increase engagement in the community and things like democracy and civic affairs, and the local news is critical to that…It has meant that we’ve been less focused on how we make money from this. Because it kind of breaks even, and that’s fine.

As we go from a search engine-driven web to an answer engine-driven web, it’s actually the local, unique, quirky content that is the most valuable content of all. If you are writing yet another story about what happened at the White House today, not a lot of value. The AI engines might read one or two of them, but by the 10th they’re like, “There’s no new information here.” On the other hand, if you’re the only organization that is covering the new sushi restaurant that opened in Park City, Utah, and what are the best things on the menu, and then somebody wants to go on a ski vacation, and they’re from Tokyo, and they love sushi…and they ask their chatbot, if that chatbot doesn’t have access to the Park Record, then they won’t know what the great sushi restaurant is, because the Park Record is the only one that’s covering that.

Some percentage of that user’s subscription should probably go to the Park Record, because that’s a valuable thing that the paper provided. And so I actually think, over time, that the more local, unique news ends up being where the value is in an answer engine-driven web. So I am actually pretty bullish that while it wasn’t the purpose of why we bought it, and, frankly, this is not a problem that I was thinking about three years ago when we bought the paper, it is turning out that the paper will be much more useful as a result of that.

Do you think that media investors should be thinking about investing in local news because of this?

I mean, I don’t know. Maybe! Somebody is going to build really good businesses around having that local news. The reason I struggle with the conversation is like, I think it’s actually hard to be a local news organization with a global parent…I think you need deeply local news.

I am optimistic that on a local basis, these things will become better businesses over time. Whether it makes sense that there should be private equity roll-ups and things like that, that I am more skeptical about.

This idea of this outcome further down the line, where publishers are being paid for their content. How long will that take to get there? And what does that interim period look like?

You’re going to see a continued decline in advertising revenue. You’re going to see a continued decline in subscription revenue. But I’m actually pretty optimistic that within a year…we will be able to facilitate substantial—and what’s substantial? More than a billion dollars, and I hope a lot more than a billion dollars—of payments from AI companies to content creators. My level of confidence on that is probably like 80%. There’s stuff that could still go wrong, but given everything that I see in the universe, given the engagement that we’ve had with the AI companies, given the fact that the next generation of AI models that have been cut off from some of the most valuable content online are plateauing and not getting better, and in some cases, the early evidence is that they’ve gotten worse.

The big thing that has to change over the course of the next 12 months for my prediction to come true is that Google needs to allow publishers a way to block them from using content for AI without having to also block them from using content for search. And if that happens or if Google just starts saying, “We’re going to pay for crawling content,” then I think every other reputable AI company will fall in line.