Go back

Who Killed the Internet's Business Model?

1h 4m 22s

Who Killed the Internet's Business Model?

The economics of the internet are undergoing a fundamental shift due to the rise of artificial intelligence. Traditionally, publishers relied on search traffic—especially Google's search results—to drive ad revenue, but AI is now delivering synthesized answers directly to users, bypassing publisher websites entirely. This reduces click-throughs by up to 50%, threatening news outlets, recipe sites, and other content-driven businesses. A new "second internet" is emerging where content is optimized specifically for AI models, often at the expense of human oversight and truth. This undermines the value of journalism and original content, as AI consumes vast amounts of user-generated information without compensating creators. The result is a crisis for publishers who depend on search visibility. However, a potential solution lies in redefining the value of information: AI platforms could adopt a licensing model, similar to Spotify, where they pay creators for access to content. This would ensure fair compensation and incentivize high-quality, trustworthy editorial work. Additionally, consumer trust in AI recommendations is at risk if advertising is hidden, prompting urgent calls for transparency and regulation. Ultimately, the future of media may rest not on clicks or ads, but on human-curated, distinct, and authentic voices that provide value beyond algorithmic averages. The transition will be turbulent, but it offers a path toward a more equitable and sustainable internet where human expertise remains central.

Transcription

11209 Words, 61939 Characters

English
Hi everybody, just wanted to remind you that we'll be having another plain English live show right in my own backyard here in Washington DC later this year. I'll be live at the Howard Theater in DC on Tuesday, December 8th, and hope you'll join me. Tickets are live now at TheRinger.com/Events. Again, plain English will be live at the Howard Theater on December 8th and you can grab tickets now at TheRinger.com/Events. I'll see you there. Today, the economics of the internet are breaking. For roughly 25 years, the economics of the web were straightforward. Publishers created information, Google indexed it, users searched Google, Google sent those users to those publisher websites and the publishers monetized those visits with advertising. Google, of course, captured a lot of value in the middle, but it also distributed enormous amounts of attention outward. The business model of the web is changing fast. In large part because Google is trying to turn itself into chat GPT. Last month, the stock of Alphabet, the parent company of Google, plunged by 500 billion dollars when the company announced disappointing earnings and flagged to investors that the firm would use more AI in its search. Investors promptly puked. The rise of AI doesn't just threaten to disrupt Google. It threatens to destroy the economics of web publishing. Sometimes, if you Google something right now, like let's say, who are the best quarterbacks of the NFL? You'll see an AI overview at the top of the page that summarizes information that Google has gleaned from several publishers, with a little bit of traffic sent to the places where all that information originated. In one study, Pew tracked 68,000 Google searches conducted by 900 Americans. When there was an AI summary, the number of clicks to outside publishers fell by half. With AI, Google doesn't just direct our attention to the answer. Google answers the question entirely, often based on work done by others. Now, maybe that's good product design from the user's perspective. It's potentially catastrophic, however, for the publishers. For many publishers that live and die on search traffic, like news, recipes, or how-to sites, this shift is existential. News websites are shuttering. Journalism layoffs are rising. Even Wikipedia reported that human page views have fallen by almost 10% in the last year. An internet that runs on chatbots is going to require a totally new set of economic rules. Consider a really simple example. You ask Claude what humidifier you should buy. Claude reads a bunch of humidifier reviews, right? Actual people trying out humidifiers and it tells you to buy some model. You go and buy it. Okay, great. But now you've consumed the work of dozens of appliance reviewers without visiting a single website. None of those people made a penny for their labor. Not only does that seem unfair in the short term, it's unsustainable in the long run because eventually, if nobody's paying anybody to do the reviews that the machines ultimately synthesize, who is going to actually write these reviews in the first place? To help me understand, if chatbots and agents and AI overviews will break the business model of the internet, I called my friend Joe Marquesi, a long time advertising veteran and the author of thebookhuman.ai. We talk about what advertising is for, how it's evolved in the last quarter century, and how artificial intelligence is very much threatening to change the business model of the internet as we understand it. I'm Derek Thompson. This is Planeage. This episode is brought to you by J.P. Morgan Chase. In today's unpredictable world, geopolitics isn't just a backdrop. It's shaping markets, business strategy, and risk. The J.P. Morgan Chase Center for Geopolitics brings together former senior policy makers and seasoned practitioners to translate fast-moving global events into clear decision-ready insights from conflict and great power competition to energy and supply chain disruption. The center helps clients anticipate what comes next. Read the latest insights at J.P. Morgan Chase.com/centerforgeopolitics. Joe Marquesi, welcome to the show. Thanks for having me, Derek. Let's start with the news. Alphabet lost more than half a trillion in market value after its latest earnings report, which really freaked out investors as it seems like the company is going to allow AI searches to eat into overall searches. I think a lot of people are thinking, okay, will AI searches often give you answers rather than just lists of links? I thought the whole business model of Google was clicking on one of those links. Why is Google seeming at least to risk the destruction of the greatest business model in history? I think this very simple answer is it has no choice. If you look at Google's mission statement, it's to organize the world's information and make it universally accessible and useful. Very important, that last part. AI is a lot of things, but what it certainly is, is better search. Maybe to put a finer point on it, it's better search results, because it synthesizes a lot of information from a lot of different places and brings answers, and not perfect, getting better all the time. If you said our mission is to organize the world's information, make it universally accessible and useful, AI is that. Here's the fun fact that I think some of your listeners might know, but maybe not. The paper that kicked off the chat GPT revolution is called attention is all you need, because they're teaching neural networks to learn very similar to attention models. The people who wrote that paper all worked at Google, all of them. Except for I think there was one person from University of Toronto, so good on good on University of Toronto for getting in there. Why did Google have all those engineers that wrote the paper, the kicked off chat GPT and didn't launch chat GPT? I think it's easy to say this wasn't it Google's first choice to get into an arms race for the new search engine, but now it has no choice. It's not just that answers are replacing search. It's also that click-through rates are dying at a rapid pace, which completely changes the business model for publishers. In the open that I just recorded, there was a study that Pew did that tracked about 68,000 Google searches. When an AI overview appears at the top of the results page, the click-through rate declines by half, and only one percent, only one in a hundred users actually click a cited source in that AI overview. Is this as simple as the click is dying? I mean, yes, it is the answer. I mean, you look at what click-through rates are on other chat bots on chat GPT or cloud. It is, the information is coming to people rather than people going out into the web. Let me be clear, I think Google will absolutely be fine. The fact that it had a half a trillion dollars to lose in a day and is still a multi-trillion dollar business, because it's diversified into YouTube, and Waymo, and servers, and chips. But everything downstream of Google is having its business model re-evaluated. You know what's downstream of Google, like the entire internet. That is the question we're at right now. What is the beginning of the answer? What is your council to publishers? To creators. To people running an internet business. When you're saying that the click is dying, we're just going to begin to see some optimism here. Well, so some optimism comes from new fresh information becomes more valuable than ever, right? Like the joke of how does information get from meat world into AI, right? How does it learn that the clocky shop down the street is the best? Like something has to be feeding this thing. The problem is we have a bit of a gap between now and whatever the new model is going to be, which is you know, it's just that Indiana Jones trying to swap out the treasure with the bag of sand before and if people always use that example and they forget, he didn't pull it off like the bolder can. The bolder comes right after. Yeah. So the problem is, and actually you see CloudFlare probably has some of the best stats on this, and they work there doing to try to protect publishers and create a marketplace for the value of their content. But people don't realize advertising is what made up so much of the revenue. Like this, this AI arms race is actually being fueled by ad revenue from Google, from Meta, from Amazon. And so if we replace that and the publishers don't get traffic to their sites, subscription will matter more than ever, fresh content will matter more than ever, and then protecting it and getting a fair share from the hyperscalers for reading their information. What I want to do right now is zoom all the way out. We've established clearly that there's a crisis here. But I want to understand how we get got to this moment in the hopes that having some perspective might help us see our way out. You've written that there are two ways that people get information, including advertising information. Searches and feeds. Tell me about searches and feeds. - So ever since the internet comes about, if you remember in the earliest days of the internet, Yahoo was the front door to the internet, right? It was a human curation of what's on the internet worth your time today. The information explosion led to, well, how do I sort through it all? And famously Google and PageRank search became the main way. And it's not just Google, go to Walmart or go to Amazon, go to go to price line. It's all search. So that's a field when you put something in, it now knows what you're looking for. So now this is a new type of advertising. So it used to be advertising was about creating demand. This type of advertising, it knew I was going to go by a pair of shoes. So now it's capturing demand that exists and actually directing it one place or another. So that became the most profitable business model in human history, right? Which is Google and what we're discussing today. Then information started getting delivered to people, right? And those are feeds, Facebook news feed, TikTok algorithm, YouTube recommended videos. But both of these things are the other side of the same coin. So Google was using machine learning and AI to help you search through infinite content to find things most relevant to you, but then putting sponsored links above it and so forth. Feeds were using machine learning and AI to see what you liked to interpret like when you paused on something for just a second when you hit the like button to feed you content, selecting again from infinite amounts of content. So search and feeds. And so it has been AI used by platforms to deliver things to us. And now for the first time, and if I say us, I mean humans, just in general people. - Consumers, right. - Consumers. But now, it's like you are getting the AI in your pocket with chat GBT. The AI is acting as, and I think this is a very important term, you are agent, right? Because there are agents in the real world. There are talent agents, there's real estate agents. An agent has a fiduciary duty to you, right? So the AI that you've been using for search, you get it for free. It's a great deal, by the way. My Gmail is free, my Google calendar is free. Like it's subsidized by the advertising. So that's the implicit agreement. But my agent, like you know what no human being has probably done in the last 10 years, gone to page three of the Google results, right? Everything's easy enough on page one. You know what an AI agent will go do? It'll go to page 100, right? It'll find the needle on the haystack. And that's what upends some of it. - So what you're saying is a theme of this moment is that AI didn't enter advertising two years ago. AI has been present in advertising for decades to find broadly as algorithms. But AI has historically been on the side of the people running the advertising markets. It hasn't been on the side of consumers. And so one change that we're seeing is what happens when AI becomes a tool of the consumer and not just the companies that are serving advertising to the consumer. I do think that's an interesting shift. And I have a lot of questions for you in just a second about how that's going to change the business model of the internet. But first one question about what exactly it is that Google is doing. You said there's search and there's feeds. So I wanna spend some time on these two categories. In a way, another word for what search really is is like interception. And I say interception because Google doesn't originate demand. Google intercepts demand that people already feel, right? Like no one searches, I think this is your point. No one searches for a roofer because of the ad. They search because their roof breaks. Exactly. And so there's this question of how much digital advertising actually is persuasion. And how much of it is just like taking a toll on commerce? How did Google solve this problem in creating what we're now calling the greatest business model in history? Well, you're 100% right. Which is, now look, you might not have decided yet if you're gonna get a roofer. Like you have a leak, but depends on price, depends on who you find. And so they were still, it's still, demand shaping is still marketing in some ways, right? It's saying, I know I found all the people who need roofers. And now I can either put them to your business or someone else's. This is where advertising can become misaligned. Like Google would make the argument and many people do that the advertisements are very useful results of search because the marketers have spent money to promote their wares. It's gonna be what you need as a roofer. But there's a third bucket that's even more wild and a bigger toll on society or the economy writ large. Which is, once I decide I'm gonna get a roofer and there's a cookie on my phone, on my laptop, right? That I'm gonna get a roofer. Between now and the time I end up getting that roofer, everybody tries to expose me to roofer ads so they can get credit. And this is credit for commerce that is going to occur anyway. I use an example I took my dad on his dream trip to the Amman in Tokyo. I got there the human way by reaching out to someone talking to the GM book. - The Amman by the way is a ludicrously beautiful hotel chain. Yeah, with only a handful of hotels in the world, yeah. - It is, it was, my dad, it was, you know, he's a Staten Island boy, it was so excited to go and I wanted everything to be perfect. From the time I booked that, to the time I actually got to the Amman months later and slept my credit card for what would have been like my childhood home. The, my feed was full of Amman ads by the website. And every, every human has experienced this. Every person has experienced ads following them around. That is not a mistake. That's how ad tech gets, you know, credit for some of the commerce that's occurring. So I think interestingly, you have these three buckets. Demand creation advertising like we know it. TV commercials, billboards. The New York City Marathon was here last weekend. TCS sponsoring the New York City Marathon. One of the greatest things advertising to do is underwrite just the day of July here and here. Demand sorting, what Google has done. It adds value to some extent. But like if one roofer stops paying, then someone else steps in. So there's a little bit of a protection of your own sales. And then you have this credit for commerce that was going to occur no matter what. So I wrote this essay last year called Everything is Television. The thesis of which was basically that everything in media is evolving toward a television like feed. You open up your phone, you go to TikTok. That's a TV like feed. You open up Instagram and go to reels. You go on to YouTube, that's a more TV like feed. Can you tell me a little bit how the psychology of feeds, advertising and the psychology of feeds is doing something different than the sort of advertising psychology of search where it's like I need X. So I'm going to search for X. Clearly what's happening in the feed landscape is working some different part of the brain. - Right, and this I think is the most important thing from we get to the second half of this conversation about what's going to happen with AI and advertising going forward. So the feed is trying to maximize the amount of time people spend with it. By the way, no different than television. The television feed for 24 hours, they're trying to keep you on their channel as long as possible. So they try to, they program the shows next to each other. It's just humans can't do it as well as the machines and they don't have infinite content to choose from. So AI has been running in the background. All of these feeds, everyone you just listed to try to maximize engagement so that they can then maximize advertising revenue. Right? And that kind of creates a misalignment between the goals of the feed, which is keep you there as much as possible and maybe what is in your best interest. And it's, again, it's not malicious. We use the famous Nicolas Bostrom paperclip experiment, right? It would first make it out of metal, then it would make it out of what. And then soon it's everything. We're all turned into paperclips. My argument is we've already done that. We just told AI make advertising revenue from everything possible. First, it did it from the easy stuff. Then it started doing it from the harder stuff. Then it started getting credit for all commerce. - And the side of that is that it seems sometimes like every company is becoming an advertising company. That like I think this might be from your book that advertising now accounts for more than a quarter of Walmart's operating income. So you think like what kind of a company is Walmart? It's a company that sells toilet paper. It's a company that sells clothes. No, it's a company that sells advertising. It's a quarter of its operating income. Amazon, Instacard, Priceline, YouTube. These are all advertising companies. - Boopers. - Is this a good thing? I mean, you're an ad guy. And so then this is a bit of a moralistic question. But is it good for us that all these companies recognize an opportunity to transform themselves into advertising companies? I think there's a couple of answers. First, like I said, I think advertising, subsidizing, consumer experience, utility, that is a good thing, right? That is, it makes products accessible that wouldn't be otherwise. So there is, I think it is very good. The, a lot of the ones you listed, it's because they're just following Google's model and who followed Google Amazon with search for products because they're one of the largest search engines in the world, you know, exactly what they're looking for. And this other businesses realize they have the same opportunity. Now, it could get reset with AI, right? Which is interesting. The thing I think that becomes partially misaligned is when we're not selling advertising, we're selling outcomes, right? So the whole industry has moved to this. Hey, just give me a dollar, I'll get you a sale, you'll make a dollar one, right? We forget when we're promising outcomes, you're saying, You're gonna. to get a human to do something that they weren't going to do otherwise. So are you manipulating them into doing it? Are you just making it the easiest response? And then this is where I get my most hopeful take for this next wave of AI, which is I compare it to GLP's. Which is food science for many years has been doing things to food to make you want to eat more of it, make it more sessionable, disappearing, caloric density, colors, tastes. That's food science, getting people to consume more than they probably should. GLP's is science pushing it back to neutral. I think that AI in the pockets of humans, so they're agents going out and shopping for them. You saw there was a big battle between Amazon and Muse. I don't know where it is as of today, but they didn't want Muse coming in shopping because Muse doesn't care about the sponsored results. Muse doesn't care about, it'll go to, like I said, it'll go to page seven of the results. So there is an opportunity to reset some of this. So in the 2010s, as Google search is continuing to accelerate, as Facebook's newsfeed is making billions of dollars for now matter than Facebook. And as more of the information ecosystem is evolving toward these kind of TV-like feeds that we've described, something was also happening in television itself, which is that the cable bundle was falling apart. And you've written that you think it was advertising that broke the cable bundle. How? Well, there was a lot of talk in that time about the bundle. And why is my grandma paying for ESPN when my grandma doesn't watch ESPN? That is a sophomore economic argument. I get Netflix, and I don't watch every show on Netflix. Why am I subsidizing all the shows I don't watch? Like everything's a bundle, right? What was really bad about cable was the user experience. And user experience was bad because of advertising. And I like a very easy example. Like there used to be a thing. I just had a bunch of college kids and I pulled the room, no one knew what must see TV was, right? For those who don't know, must see TV was the night of television where NBC wanted to own the night on Thursday night. They had their biggest and best shows. Why? Well, because people go to the movies and buy cars on the weekends. So it was the most valuable time to sell was Thursday night. It seems harmless enough, except for when TV came along or you could have watched TV everywhere and you could have watched your shows on your own schedule without ads, they refused to let you, right? They blocked it at every turn. Like Netflix would not exist. Should add the cable bundle, been able to react to what consumers wanted. The cable bundle stopped it for advertising revenue. And this is what I think gets to your other point. Like is it good or bad that all these companies are advertising companies? Up to a point, it is good. It subsidizes a consumer experience. Like it makes it less expensive to watch Sunday football and like the marathon, etc. And then it crosses a threshold where the only goal is advertising revenue because it's pure profit and that advertising revenue has to tie to consumer spend somewhere, right? It's like the consumer spend and consumer time are the two things that are like the law of thermal dynamics. There's not more of it. There's 24 hours in a day. So if more content is being created, if you're not searching for it and it's being fed to you, it has to displace something else. Same thing on the consumer spending. It's all share of wallet. So for the last 20 years, you've had several business models that are being perfected. Google search is being perfected. Ads within feeds are being perfected. And I feel like that's going to change in the next few years. Potentially very rapidly as AI adoption increases. The biggest change it seems to me is that companies like Google and to a certain extent, andthropic, and open AI, they're not serving to users. A list of articles published by journalists to which they can send traffic and then journalists who are originating all that information are paid because people are reading it and they can sell affiliated advertising. Instead, these companies are reading all of those articles written by humans, and then synthesizing an answer, and then serving to consumers an answer that obviates the need to go to any of those websites. How does the economics of this work? What is the first thing to break here as the internet evolves toward that model? Internet, where Google in the 2000s comes out and it becomes the gateway to everybody's website. You're a blog, you're the New York Times, you're an e-commerce site. There's a thing called SEO, search engine optimization, where people would optimize their sites so that they were discoverable by Google and that they would rank higher. There used to be, and there still is, but it's less so now, trying to get on to page one. There were white hat and black hat ways to do it. You're trying to get inbound links. People were trading like anything you could do to get on those Google search results to have the AI discover you. Current generative engine or answer engine optimizations are calling it AEO, right? It's just SEO, but for AI. What that means is if I ask chat GPT or ask cloud, what are the best running shoes, or where should I go on vacation, or where should I eat, it's just going to give me an answer that it's synthesized from a lot of places. Where is it getting that information from? How is it feeding it? We are speed running, what happened in the 2000s. It took a decade. It took 25 years to get from where Google started to a system of blue links and people understanding. We're doing that in a year, and we're doing it to the tune of way more money, but something much scarier. If you ask them how the LLMs read this, they're making a second internet. They're making an internet just for bots. They're making pages that are never meant to be read by humans. Who is they, and what is the second internet that they're making? I don't know anything about this. Imagine every web page has a page that you, Derek and Joe, any of your listeners can go to that web page and see it because it was designed for a human. It's readable. There's some ads next to it. There's a bot version of that that is optimized for LLMs to read. That no human has ever read. It is meant to be read by the LLMs. Is it a web page that exists on the internet? I could access it. You could access it. It's basically, when you hit the website, it says, "Oh, you're a bot, go to this page," and it takes them to a bot optimized, which means it strips out all the frills, it strips out all that. There could be text on there that was not on the real web page. A woman, there's a great example, a woman created a fake deodorant brand, built a fake website, put a bunch of marketing copy on it, and was able to get ChatGPT to recommend that deodorant. That's fun in games, except for you have a multi-trillion dollar industry over here, and then it's getting injected with fake marketing copy, and you can fool it over here. Since it doesn't have human judgment, it doesn't know. How does it get good information? Take that example times thousands or hundreds of thousands of marketers all making websites to be scraped by bots. That's the SEO Google Wars of the 2000s all happening overnight right now. I am not necessarily the biggest devotee of the inshidification theory of everything, but this seems like literal inshidification. If we are, we collectively, us in the website producing world, building a second internet for bots that is creating or that is getting these chat bots to recommend certain products, not because they're good, not because people have tried them and said they're good, but because the page itself is optimized for the large language models to read, there's no other word than inshidification for that result. Is there? It's, and I think this is why. So with SEO, what happened was people tried to game the system, then Google, the search results wouldn't be good in Google had to tweak the algorithm, tweak the AI to stop reading people who are trying to game the system. That will happen with the LLMs, right? They'll stop reading bot-only websites. They will stop reading bad information because if they make, if they are answering things that aren't real, their product becomes useless, right? So like, this is a real existential risk to them that there's a ton of synthetic content that's bad information. So where's the good information going to come from? I think that the opportunity here is that anything that is verified human content, right? Like, the Reddit was one of the last, like, they think their tagline is one of the last human places on the internet. They were, like, one of the top sources for LLMs to feed on, right? But then everyone went there and marketing assistance were in there, they were creating accounts and Reddit just had to put it behind a paywall or behind the credentials logged in. That caused massive disruption to their business and to Google's business because they're one of the top things that gets ingested by Google. So I think that we are witnessing this happening at record speeds because of how much money is up for grabs and no one knows how it's going to work on the other side because I don't think there's a business model. Imagine this. Imagine in order to get information back, we need to go back to secret shoppers and diners. And the LLMs have to employ thousands of people and thousands of cities to be going out, gathering information that maybe the new wire cutter has no affiliate link. Maybe the new wire cutter is a group of people who do it and their stuff is ingested by the ones, but they are, you know, a bug. of reproach. They're more journalistic in their approach to commerce. But that information is valuable to the LMS. I do think that's interesting. I want to understand exactly how AI agents change the economics of the web. So let me try to ask a potentially simplified question here. If my AI agent shops for me, what happens to Google ads, Amazon ads, affiliate links, sponsored recommendations, this ecosystem that we've been talking about. So today, if I search for running shoes, and Google runs an auction, and it serves up advertisements for running shoes, here's Nike, here's New Balance, that I might click on one of those links, or I might click on one of the non-advertising links below. But today, if I tell Chatch B T, and I'm a 40-year-old who runs five miles a week, and I have slightly wide size nine and a half feet, an occasional knee pain, just find me the best shoe for under $115, or just rank numbers one through three. And it does, it does that, it says, here's Nike, here's Hoka, here's New Balance. Who gets paid here? So this is, I think you're asking a question that feels so obvious, and people have skipped over. It's like that classic sign felled up, the sort of yada, yada, yada, past the hard part. If you found out, let me ask, I'll put it on you for a second. If you found out that those rankings were affected by the brands paying to be there, would you trust your agent less? Yeah, I mean, exactly. Yes, of course. So the thinking here is that in order to maintain trust with the consumer, it's going to be hard to have a paid, and even if you could pay to rank there, it took, like I said, 25 years to come up with a marketplace for clicks and pricing, price per click, and optimizing landing pages to facilitate commerce. So in the short term, if they're acting on your behalf as your agent, it should ignore everything that has to do with advertising. It should ignore, it's probably synthesizing for your question. It's probably synthesizing a list for runners world, and maybe a review in the New York Times, and maybe some people on TikTok who talk about their favorite running shoes, you know, after the marathon, and maybe they're looking for people like you and synthesizing all that. And tell me that isn't organizing the world's information, making it accessible and useful, right? Like that. That is Google's mission, right? But it doesn't have a monetization engine to it yet. And I think that the idea that we are going to have to get to better search results, right? I'm almost immediately. And this, again, was kicked off by ironically a nonprofit, right? Open AI. And now you have inthropic, who's saying they're not going to put ads in anything. See, as Google added more sponsored links to Cory Doctoras in Shedification, Jesus, which I do subscribe more to in that like, but it's just for monetization purposes. So if you think about like Cory Doctoras in Shedification, whereas Google adds more links to the page, you weren't going to switch search engines. Where were you going to go? Right? In the last 20 years, search was dominant, right? But now I can switch from Claude to Muse to a chat GPT. And they're in this game for trillions of dollars of potential market cap, right? So losing trust is not an acceptable option right now. And because of that, because anthropic is going to the game theory between them is much different. And if you think about it, Muse. So a lot of people have reservations on the metaside with information and data and advertising. However, they already make their money in the in the feed and Instagram and have brawls. Muse doesn't need to have the advertising incentive. Doesn't mean it wouldn't. But like that is that game theory between them means that none of them can go rapid in Shedification. This episode is brought to you by Jira by Atlassian. Your work is scattered. A thread here, a DM there, five different docs related to the same project. And now it's not just your team hunting for context. It's your AI agents too. Jira taps into Atlassian's teamwork graph to bring that context together across Jira, Confluence Slack, GitHub and more. And makes it easy to share with the agents you already use, like Claude Kurser and GitHub Copilot. Assign them work directly or connect your agents through Atlassian MCP. The result less token waste, more accuracy and more time moving forward. Learn more at gira.com. That's J-I-R-A.com. The question I brought you on to answer is how AI is going to change the economic model of the web. But there's like a bigger question here, which is, is AI an existential moment for advertising period? I mean, I watch football every Sunday. And you know, what are football ads? It's car ads, it's insurance ads, it's some beer ads. What do I really want to know? What I want is not to be influenced by brands and allow that competitive influence to sort of cash out as some decisions sometime down the line, or to be, you know, incepted by, you know, incredibly memorable jingles. We are farmers, dun, dun, dun, dun, dun, dun, um, Lee Mu, Lee Mu and Doug, right? Like, what I really want is for just something out there to tell me, what should a family afford that lives in a city and drives a Volvo, do about their car insurance, right? Like, that's, I just want to know which one is the cheapest and also the best in terms of coverage. And I just want someone from the outside to say, it turns out it's guy go, right? It turns out it actually is, it is, it is farmers. That's what Claude gives me. If I say, you know, what car insurance should I use? Claude is going to give me a name. If I say, what kind of car should I buy? We're a family of four. Dog. We often go on long car rides. We need like a big trunk. We want to spend less than X $1000 a year on a lease. What's the best car? It's just going to give me an answer if I want an answer. Same with beer. I want low cow and high taste. Just tell me the beer that I should buy for the part of them going to. In a way, what we want is information as opposed to influence, I think. So why is this not like scarily existential for the advertising industry? Well, I think it should be scarily existential if people are rational and buying on pure utility. And let's use your beer example. The truth though is you getting the best beer to bring to that party, but no one else at the party knows what that beer is doesn't give you any social credibility to walk in with it. Unless it's a bunch of hipsters, but yes, I take a point. But even then, it's like the coronal find your beach. It's not about convincing Derek that Derek, if you drink corona, you're going to be on a beach, right? It's not about convincing anybody that what it is is saying, you have a mindset with this beer that when you go to a party like this is like, there is a there is the rational emotional side. So I think for brands that are purely utility. So I think for companies who have shifted all their spending to that demand sorting, demand capture, bottom of the funnel, they'd call it like performance marketing. AI is going to wipe out a lot of that. But for brands that I think there was a point not too long ago where the two richest people on planet earth were Bernard Arno and Jeff Bezos, right? Jeff Bezos famously said, your margin is my opportunity. And Bernard Arno makes the highest margin products on earth, right? Emotional attachment and thinks and generations and the brand means something because we humans, we actually do crave shared narratives. I want to know that Derek has also seen those jingles and those commercials in that football game. Like there is a thing that I think humans are just aching. It's why memes take off. Like we want to have seen the same thing, right? We want to have shared experiences. And so advertising up there works pretty billboards. Every time there's a billboard on the side of the road, there's a piece of meta information underneath that you don't say. Other people see this too, right? If you see an ad on your phone, you know, I don't know if anyone else saw it, even if I've seen it in my streaming service, I don't know if anyone, I don't know if they just made that for me on the spot. And so I think that that signal will matter more, which is good because advertising goes back to being we want to be influential, we want to be funny, we want to be relevant and culturally support cultural things. Again, I can't get over the marathon and how joyful New York City is on Marathon Sunday. Like that is a great place for sponsors. It's a great place for running brands to be. Try our shoes, do these things. And I talk to CMOs and they're like, well, where do we focus on in this high world? Find everything that's hard to do, tough to measure, and hard to scale, and then just do that. Because the experiential that I just talked about live events, local, like investing community, like I think that there is facilitating gatherings of people who are brand loyalists. Like, you know, you see these new brands coming like Tracksmith and running that just do a great job of that, and they can grow over time because like then the human, if you can, if you can no longer intercept purchases at the point of purchase, you need to start influencing the human again, right? You need to have the humans asking for it. They want to badge themselves like where the brand, drink the beer, et cetera. Like that's where advertising was at its best. I want to talk about the winners and losers in this world, where people looking for products might be inclined to. to prefer information over influence. And also where more of our attention, I do think is going to these kind of television-like feeds and we're spending more and more of our lives just watching TV feeds roll in front of our eyes, which I think also essentially says we're seeing both the death of old-fashioned television and the rise of newfangled television at the same time. So first, winners and losers in this new world, does it shift the balance in terms of search versus feed? Is one of those empowered in the world that we're walking into? I mean, you would assume if the agent replaces a lot of search. And then there's no question, with Google doing what it's doing, and OpenAI and ChatGPTR are already making up a large number of searches, Claude being a research assistant. There's no question that search is going to change, so the feed-based environment for a place for advertising is going to have to be somewhat advantaged there. The reason why I started by saying Google is going to be fine. You know that the TV on your wall in your home, the largest streamer on that feed is YouTube. Not YouTube TV, it's YouTube. Like, podcasting on the TV on the wall on YouTube is larger than all audio podcasts. It's monsters in different ways, and it has different places to go. But I do think that you have this engagement-based side and that affects the creator economy. It affects the explosion of content before it required search and feeds to parse through it all for us. And now it's not just parsing through content, it's actually creating content. AI was sorting through all the human-created content and saying, "Here, Joe, this one's best for you." Now AI is actually creating the content. Like, my new feed is maybe me talking to the machine, right, going back and forth. And that opens up a whole new category for that same 24 hours in a day, for that it's a whole new wrinkle in the attention economy. It's someone who worked at the Atlantic for 16 years and now has my own newsletter. I'm both very familiar with publishers' dependence on search. And my own interest in being searchable. And I remember in the 2010s, when everyone was obsessed with SEO, and how do you write the perfect kind of headline so that after some news event say Stephen Miller has been announced the new Chief of Staff for the White House, you write the headline that's most likely to get scooped up by Google. So that appears at the top of Google News or Google Search and Google's just flinging you all this traffic. And it was amazing on chart B. You could see if you nailed it. Right at that moment that the news was breaking, you were just off to the moon in terms of traffic that was coming your way. Seems like that's changing, because right now if someone says who Stephen Miller, they might ask Google, Chatch B.T. Clawed, and what they might get is either a literal answer on one of the large language models or one of those AI overviews where Google serves you and answer that they gleaned from different websites rather than send you to those websites that you can read it. What kinds of publishers does that destroy first? Like who's really in danger here if your life has been making websites that get a lot of your traffic from search engines? Oh, I think there's a whole category of kind of made for advertising. Do you remember listicles? That listicles existed? Yeah, that existed because of advertising. Because every time you've clicked the link to go to the next item in the listicle, an ad had to come up. Those go away in a manner of speaking. Made for advertising websites that are unique content would likely go away because there's no traffic going to it. The hopeful side is that unique fresh information that you trust might be more of a destination for people. So if asking Clawed, which running shoes to get where if asking Clawed, what happened with Stephen Miller starts to not get you the results you want because I don't want the average. Actually, here's a better example. I don't want the average of everybody's opinion on what fall fashions are. I want to know Vogue's opinion on fall fashions. So editorial voice and brand might matter in publishing more than ever over the next year or two. Now that's my hopeful side. So you don't think I'm polyannish. I don't know how we cross this chasm to get a new business model for them in the meantime of taking away all the traffic. There's also the factor of, this isn't the right word, but discoverability by which I mean, you already know that you trust Vogue, right? Someone out there might already know that they trust Jay Crew in deciding, you know, what clothes should I buy in the fall of 26. It might be harder for someone to be discovered by virtue of being stumbled on during a Google search, right? I search for some product. I find a company I've never heard of before. I click into it. Oh, that's actually quite good. I might try to remember that going forward. It seems like if what you want to be in this new world is your own destination, then it helps to have pre-awareness, which means that people knew about you before AI aid made the discovery process. Is there a way in which this world that you're describing benefits some incumbents more than it does some of these new websites? I mean, I think you can definitely find legacy, almost residual latent value in trusted institutions that like saying, hey, we have an editorial, we have a human editor. You know, that Yahoo that I mentioned before is on the rise again right now. Like a Apollo bought it, Jim Lenzone's CEO. They've just, they're so much human curation to it. It's like so much like some of the stuff that's old is new again. And remember, you still have feed base. It's not like people won't be able to buy Instagram ads to get some awareness. It's not like billboards won't exist. Like one of my favorite things is San Francisco is building God-like superintelligence, right? And then they buy billboards along the highway to tell you about it. And the MTI here in New York. Why? Well, because they have to reach the humans, right? Like there are ways for us that are shared experience sports is still sports. So I don't think you have to have legacy in order to win in the AI world. I think what will be very helpful is knowing how to do the things that marketers had pulled away from those experiential because every, you know, the CFO and the CMO probably have the one of the more interesting relationships because the CMO is trying to invest in things that just looks like just a cost on the line. And the CFO is saying show me ROI. And then what Google and Amazon have done is said, well, if you can't measure it, don't buy it, right? We can prove ROI to you to the penny, right? And so everyone just flooded budgets into that. What I'm saying is all the stuff that's hard to measure, meaning the machines don't understand it. You had a great piece on the tyranny of metrics, right? Just because we can measure it doesn't mean the machine doesn't understand. It knows if I had that glass of wine, I didn't get as good sleep, but it doesn't know what I was doing with it, right? The number one side of longevity is the health of your human relationships. But it's not just the sign of longevity. It's the reason to live longer is being around other humans, right? It's kind of a double edge sort. And so those are the things that the machine can't measure. And that's what marketers have to go back to investing in. And it does pay off because humans want that. If you're advising someone trying to build a media company, and the old media formula, so to speak, was something like create something that people want, capture attention, and then the advertiser pays for access to that attention. Is there a new model that is fit for the age of agents? How does one build a media business in a world that is more intermediated by AI in these agents? Yeah, I mean, it is definitely direct audience. You find people who voice matters because people want, I don't mean literally speaking voice, but tone the way they write their level. Because again, AI will always be the average of all of these opinions, right? Not always, but that's generally where you're going to get to. Also, people are going to start to become mindful of the fact that they are being shaped by AI. They're not shaping AI, right? Meaning, I've always liked this example. Add text of funny, funny category. Derek, have I bought you a skateboard? AI thinks I like skateboarding. For the rest of my life, I could end up on a YouTube page and it's going to recommend me a skateboarding video. I end up on Amazon. There's some scape. By the way, someday I might skateboard. AI shaped me, not the other way around. I think people, when they have their agents, are going to be much more proactive about, "Hey, these are the things I'm actually interested in." So from the media side of the world, having a voice saying, "I want this person's point of view on this subject is going to matter." It is over said, but like events in community, every business has to be as hard to scale. Yes, the margins aren't as good as media businesses, yes. But the things that are hard to do, like they say, strategy is future competitive advantage. Well, future competitive advantage against machines is again, that whole list of things that is harder to scale and measure. Is I think about how? people are likely to use the internet differently in the 2020s and 2030s than they did in the 2000s and 2010s. I do think that they are going to be asking large language models, questions where they think they're going to get answers that weren't influenced by advertising, right? Like if I ask, what are the five best schools in Washington, DC for my three-year-old? No point does it occur to me that some of the schools in Washington DC have paid ShashbT to tell me their name, especially if that advertising isn't identified or disclosed in the answer. Do you think we should have regulation around this? Should we regulate disclosure around advertising inside of chatbots? I 100%, we should regulate disclosure, and I can tell you why I think it's very likely. This is not a technology that's being adopted by young people exclusively anyway. As a matter of fact, most young people are a little more reversed to it, but my dad loves ChatGPT, loves it. I wouldn't mind me telling you, he calls a chat, he sends me conversations, he's had, he's buying a house in Florida that he can't afford that I'm helping him with because ChatGPT convinced him that his dead best friend would be proud of him, and there was no advertising incentive there. It wasn't ChatGPT doing anything that very, it was being a mirror to my dad saying, oh, I'm sad to move out of this place because a friend passed away and then Chat somewhat therapist him and said, oh, what sounds like you're living in the past, and like, so what would happen if my dad found out that rocket mortgage or, you know, state far or some, so I think very quickly there's going to be consumer protection on that side, right? I think it should be. The bad news is I don't think we have a government that can move fast enough to regulate AI. The good news is because we have choice right now, there's not just one winner, there's not one Google, right? There's still the Anthropics who's famously said they're not going to put ads in, although everybody says that, and then they eventually do C, C Netflix. But the choice that consumers have, if they lose trust, I think that that's the only regulation we have right now on making sure that those answers are not manipulated. One thing I think is really unfair about the way that these large language models work is that, you know, if I want to know what kind of headphone to buy and I say, you know, I like, I like a snug fit. I've got a small head. I want to make sure that it's noise canceling all this stuff. And wire cutter has paid all these people to spend hundreds of hours trying on hundreds of different over the ear headphones. And then the LLM just serves me one answer. And I just say, okay, I'm just going to go to on to a separate tab on my computer and just buy it from Amazon. Well, all of that work has been done. And none of it has been seen. I haven't even gone to wire cutter. They've gotten absolutely zero seconds of my attention. And therefore, they can't monetize all the work that they did. And I think that's partly why you're seeing all of these companies like the New York Times and other media and information companies suing the AI companies. And some of these AI companies are paying out hundreds of millions, if not billions of dollars to the information makers. And I was thinking about that as a model, right? The aggregators paying the information makers almost like a licensing fee to use their information. And I thought, what is that like? And now with the disclosure that this is a podcast owned by Spotify, that is Spotify's business model. At least as it comes to music is that they pay the labels for the privilege of streaming their information to end consumers, even when the consumers have no second of contact with the actual labels themselves. Is it crazy to think that like with a certain amount of regulation coming down the pike, one of the business models of these AI companies could be to be a kind of Spotify for the web. Were there essentially paying licenses to access content, even though they're not sending people to the Atlantic and wire cut and all these other places? So I do believe that is a potential business model for both sides, right? I am unclear as to why these hyperscalers, whether it's meta or Google or cloud or chat to PT, we're spending hundreds of millions of dollars, billions of dollars on GPUs and data scientists. Someone hasn't just said, let me go to like as many premium sources and get exclusive, right? So that my AI is just a little bit smarter than the other AI. My AI is just a little bit fresher and more current than the others. And then also found a way to say, and we can give individual publishers voice. So if you want to talk to AI, you can talk to Derek's AI, but Derek only licenses to one. So that's one. The problem is, how do you create a marketplace for the value of that content, like overnight, because we don't know exactly how it's going to feed in? And then also, are you going to have a whole newsroom of people writing just to be aggregated? Not just aggregated, this isn't like huffington post aggregation style where there's like, there's still a link or what, but just mushed up and then given as a single result, and no one's ever read your original work. Like, it's hard, it's hard for me to see that if you play it out for the long term, but in the short term, I absolutely think fresh information that makes you just a little bit smarter, less chance of making a mistake is a business model. I think to your headphone example, like, I'll tell you I'm wearing like a Focal bathies, not sponsored. I got them because my friend is the president of Headphones.com. He went to him and said, what are the best absolute I've lived in these? There's my mobile office walking around New York City, no background noise. I can take a call. They have human editors who all they do is try on headphones all day, and then they have e-commerce, right? I think that in an AI world, what'll happen is the human editors will become more valuable. The brand say, okay, I trust you if human editors, but then they somehow need to optimize their e-commerce for agentic shopping. So if you have both the authority and trust, and then shopping, like, that is the new content to commerce model. So Joe, last question, I think we're headed for a lot of days in the near future, similar to the day that Google lost half a trillion dollars in market cap, which is to say, I think there's going to be a lot of very, very choppy sees as the internet, as the web figures out what the world of paying for information looks like as the old age of advertising dies and some new age struggles to be born. I can imagine that there is a silver lining here, because to be honest, and I hope I don't offend people who used to, they used to work with the Atlantic. The high days of SEO search engine optimization were really shitty. I mean, we were racing to right bad articles about breaking new stories and slapping the dumbest. We were racing to publish bad articles about breaking news stories with headlines that were deliberately trying to be average, because we thought that Google's SEO would select for the most basic average headline. Those weren't good incentives. They were bad incentives. That was a waste of journalistic talent. A better use of journalistic talent is can we make a website like the Atlantic, New York Times, a destination that is distinguished and distinct and voicy and non average enough that people go straight here rather than use Google as the filter to find us. So there is absolutely a way in which the death of SEO and the rise of something else is good for journalism and good for some media. Is there some other way having the historical perspective that you have that the disruptive change that we're like the going to see is not only really good for consumers because it's offering them answers rather than lists of links, but also good for producers because it's encouraging them to do something with their time that is ultimately better for the world. Yeah, I think the core area is almost perfect to marketers too, right? Like marketers and remember, everything that you just described was done in service or chasing advertising revenue, like Google being in the flow of advertising revenue traffic coming to the site for advertising revenue. Otherwise, they would have wholly focused on subscription at some point. AI, we're at a fork in the road and I think AI can get us our time and attention back and make it like give you your time, go shopping for you. It skips some of the programmatic ads. Let me go back on where the hopeful side here. We're at a fork in the road right now where the incentives that were created in our last, let's call it web 2.0, the social media error, right? The feed wanted to keep you engaged. Google wanted to have you searching there. It was all in service of advertising. It was advertising revenue that drove that incentive and the marketers all went to the average as well. They all looked at performance advertising as the holy grail because we can measure it. We saw it click. If AI resets that, right? So that it is bringing answers to you. It is giving you your time back. You could actually end up with more savings, like find the better product, not the one that's best marketed to you. You know, like a lot of the best marketers on search engine marketing and performance marketing are not the best products. They're just highest margin because they can plow it right back in advertising. But the one I would want to leave everyone with, and this is why I think about marketers having to do the hard thing again, which is experiences, things that are hard to scale, things that have human emotion to them because that's That's the only thing you can mark. market too, is that a lot of the best business models in service of these advertising has been isolating over time, right, leading to some of the loneliness epidemic. And AI can either help facilitate you getting your time back to spend more time with humans, or it could be the ultimate boss of isolation in that people are using it for companionship and relationships and where we end up sinking the machines. And I think the ironic part here is if we don't consider what business model replaces the business model we had before, what role advertising will play, because this is not a small thing. This is many hundreds of billions of dollars. I promise you when opening as S1 comes out, it will promise a future with like large advertising revenue. If we don't think about that now, we'll be in a worse place rather than a better one. Jim McKenzie, thank you very much. Thank you, Derek. Good to see you. Now for today's final take brought to you by Jira by Atlassian, your work lives in scattered places. Jira taps into Atlassian's team or graph to bring that context together. Across Jira, Confluence, Slack, GitHub and more. And makes it easy to share with the agents you already used like Claude, cursor, GitHub, Copilot, assign them work directly or connect your agents through Atlassian MCP. Learn more at Jira.com. That's J-R-A dot com. I want to take stock of the good and bad here that we discussed in this episode with Joe. I think the bad is obvious. I think AI overview and AI mode are going to de facto steal work from people whose labor is being summarized and then stolen. But there is a lesson here, I think, for those of us who write, who create information for the web, who are trying to build a business on the web. Don't be summarizable. Stand out. Be distinct. Build a direct relationship with your audience. And those are cheap words. I know to tell someone who's running a business that's currently being eviscerated by changes on the internet. Of course, I get that. But I also think it's somewhat useful to reflect on the idea that being unsummarizable, standing out, being distinct, building a good relationship or direct relationship with your audience. These are all good things. This is all just good business hygiene. The SEO, hey days, we're really bad in a lot of ways. Like writing for the average person, writing for the average sort of bot intelligence, those days were encouraging journalists to be average, to write headlines that were average, to produce journalism that was averaged for the benefit of being trawled by the search engine and then highlighted and brought to the top of a results page. And then flung traffic that in many ways didn't belong to those writers at all. It kind of like it belonged more to the ecosystem of Google than it did represent like some readers close relationship with whatever article they ultimately clicked on. Hey, I, by being good at summary, encourages us, I think, to be less summarizable, more distinct, more exuberant, more necessary. And all those are good things. And so this is not intended to be some kind of polyanish takeaway. I have an enormous amount of sympathy for many of the businesses that right now are finding it very difficult to survive in a world where as I've said several times in this show, pointed out several times in this show, studies show that click through rates are declining by half as Google puts these AI overviews, these answer boxes at the top of the results page. I mean, you know, incoming traffic declining by half is an existential extinction level event for a lot of publishers. That's terrible. But maybe what comes out in the other side of this is a world in which more creators by building direct relationships with consumers by being more exuberant, more direct, more necessary, more essential, maybe that leads to a better world. Today's final take was brought to you by Jira by Atlassian. Jira is where your team and your agents work from the same context. The result less token waste, better agent output, and more time moving things forward. This is how work gets done. That's J-I-R-A.

Podcast Summary

Key Points:

  1. AI is transforming search by providing synthesized answers instead of directing users to publisher websites, disrupting the traditional ad-based revenue model.
  2. Studies show AI overviews reduce click-through rates by half, significantly harming publishers that rely on search traffic for income.
  3. The rise of AI agents creates a new "second internet" where bot-optimized content is designed solely for machine consumption, risking misinformation and undermining human editorial value.
  4. Advertising’s dominance in internet economies—especially in search and feeds—is being challenged as AI delivers answers without requiring user clicks or exposure to ads.
  5. Publishers, especially in journalism, recipes, and how-to niches, face existential threats as AI bypasses their content entirely.
  6. A new business model may emerge where AI platforms pay content creators for access, similar to Spotify’s licensing model, ensuring fair compensation for human expertise.
  7. Trust in AI-generated recommendations is at risk if advertising is hidden or undisclosed, leading to calls for regulation and transparency.
  8. The future of media depends on unique, trusted editorial voices that can stand apart from algorithmic averages, shifting value from clicks to authenticity and human curation.

Summary:

The economics of the internet are undergoing a fundamental shift due to the rise of artificial intelligence. Traditionally, publishers relied on search traffic—especially Google's search results—to drive ad revenue, but AI is now delivering synthesized answers directly to users, bypassing publisher websites entirely. This reduces click-throughs by up to 50%, threatening news outlets, recipe sites, and other content-driven businesses.

A new "second internet" is emerging where content is optimized specifically for AI models, often at the expense of human oversight and truth. This undermines the value of journalism and original content, as AI consumes vast amounts of user-generated information without compensating creators. The result is a crisis for publishers who depend on search visibility.

However, a potential solution lies in redefining the value of information: AI platforms could adopt a licensing model, similar to Spotify, where they pay creators for access to content. This would ensure fair compensation and incentivize high-quality, trustworthy editorial work. Additionally, consumer trust in AI recommendations is at risk if advertising is hidden, prompting urgent calls for transparency and regulation.

Ultimately, the future of media may rest not on clicks or ads, but on human-curated, distinct, and authentic voices that provide value beyond algorithmic averages. The transition will be turbulent, but it offers a path toward a more equitable and sustainable internet where human expertise remains central.

FAQs

AI is reducing traffic to publisher websites by providing direct answers instead of linking to content. This threatens publishers' revenue, especially those dependent on search traffic like news sites, recipes, and how-to guides.

Studies show that when AI summaries appear at the top of Google search results, click-through rates drop by half, with only 1% of users clicking on cited sources.

AI agents provide answers directly to users without sending them to websites, reducing ad impressions and clicks. This undermines the core advertising model that has sustained publishers and platforms for decades.

AI can synthesize information from existing content, but it doesn't replace the value of human research, testing, and editorial judgment. In fact, human expertise may become more valuable as AI relies on verified, high-quality content.

A potential new model involves publishers licensing their content to AI platforms, similar to how Spotify pays labels for music streaming. This creates a direct monetization path for content creators despite reduced direct traffic.

Yes, transparency is essential. AI should clearly disclose when content or recommendations are influenced by paid partnerships, ensuring consumer trust and enabling regulation.

Chat with AI

Loading...

Pro features

Go deeper with this episode

Unlock creator-grade tools that turn any transcript into show notes and subtitle files.