For most of the internet’s history, publishers had a familiar bargain with platforms. They supplied the information, the platforms helped people find it, and some of that audience came back. AI answers have broken that arrangement. Publisher content now gets summarized inside a chatbot, often without a single click, and licensing deals are how some publishers get paid for it. The AI platform still owns the interface, the data, and the relationship with the person asking the question, which leaves an open question about what those deals are really worth.
Listen or watch:
Rob Kelly may have the best-kept spreadsheet in media: 99 AI content licensing deals, going back to The New York Times’ little-discussed $100 million arrangement with Google and running through Yelp’s new deal with OpenAI. The creator and host of Media & the Machine came to the subject as an operator. Early in his career he ran the content licensing business at CMP Media, selling a feed of a million magazine articles to LexisNexis, Bloomberg, and Dow Jones. He describes that business as close to pure profit, and the experience shapes how he reads today’s deals.
Kelly and host Pete Pachal start with the two paths that defined the early market: the Times suing OpenAI at the end of 2023, and News Corp signing a five-year, $250 million deal with OpenAI a few months later. Kelly’s read is that the Times sued quickly because it already had a playbook. It had licensed its content to Google and knew what it was worth. News Corp’s deal works out to about $50 million a year, which sounds modest next to a media company’s revenue. As profit, Kelly argues, it could be bigger than what most of News Corp’s publications earn.
“This is not normal revenue. This is gravy, highly profitable.”
Then the market shift. In 2024, the vast majority of announced deals included training. Today, by Kelly’s count, training is the emphasis in roughly 40 to 50 percent of deals, with inference and retrieval catching up. He doesn’t think training is finished. More of the world gets recorded every year, from phone video to the cameras at Madison Square Garden, and he says anyone who claims the models have run out of content to train on is mistaken. The money, though, is moving toward what happens at answer time.
The most useful idea in the conversation is what Kelly calls licensing content to commerce. Yelp’s deal with OpenAI includes a request-a-quote button, so a search for a plumber can turn into a lead for Yelp inside the AI answer. HubSpot, as Jonathan Hunt told this show, cares less about how much of its content gets scraped than about the subscribers that flow back. Kelly expects the Times to settle with OpenAI on similar terms, with subscriptions, games, and recipes sold inside the answer. If the back end is strong, the scraping matters less.
He also suggests publishers stop measuring AI traffic against what Google used to send. A new business, he says, would read the same numbers very differently.
“If you and I were to form a startup tomorrow and we weren’t getting any traffic from Google, but we’re getting traffic from ChatGPT and Gemini and Claude, we’d be thrilled.”
The conversation also covers the mid-tier squeeze (big publishers have lawyers, small ones mostly want to be discovered, and the middle has neither), ProRata and why, as far as Kelly knows, its publishers haven’t been paid yet, the RSL collective licensing protocol, the idea that Google or Apple could import the YouTube revenue-share model, the backlash against AI among younger people, and Kelly’s 8.7x productivity number, which he measured in a spreadsheet. He closes with two futures for media, one where it loses and one where it wins, and a prediction.
“I believe that OpenAI will settle with The New York Times. They’re going to want to settle before their IPO.”
Kelly expects that deal to be transformative for the industry, and he hopes the Times’ stand gives other publishers something of a playbook.
In this episode
Pete and Rob discuss:
- Kelly’s path from Wall Street and tech journalism to running content licensing at CMP Media
- The two early playbooks: the Times sues OpenAI, News Corp signs a $250 million deal
- The Times’ little-discussed $100 million licensing deal with Google
- Why licensing revenue is close to pure profit for publishers
- How deals have shifted from training toward inference and RAG, and why the models haven’t run out of training data
- Whether AI companies pay mainly to avoid lawsuits
- Licensing content to commerce: Yelp’s request-a-quote button and a subscribe button inside the answer
- Google Zero and taking a beginner’s mind to AI traffic
- How the AI answer is being renegotiated, from ads for bots to brand links
- The mid-tier squeeze, ProRata, and the RSL collective licensing protocol
- Whether Google or Apple could import the YouTube revenue-share model
- The backlash against AI among younger people, and an 8.7x productivity gain measured in a spreadsheet
- Two futures for media, and why Kelly expects OpenAI to settle with the Times before its IPO
About the guest
Rob Kelly | Creator and Host, Media & the Machine | LinkedIn | Media & the Machine
About the show: To explore more conversations like this and see what’s new, visit the Media Copilot website at mediacopilot.ai. You’ll find new episodes, expanded resources, and tools designed for journalists, communicators, and media leaders navigating the fast-changing world of AI. It’s the home base for everything Media Copilot and it’s just getting started.
Enjoyed this episode? Subscribe to The Media Copilot on Substack, Apple Podcasts, Spotify, or your favorite app. On YouTube? Tap the Like button and Subscribe to the YouTube channel.
Produced by The Media Copilot.
Music: “Favorite” by Alexander Nakarada, licensed under CC BY 4.0
All rights reserved. © AnyWho Media 2026
TRANSCRIPT
[0:02] Pete Pachal
Hi, welcome to the Media Copilot. It’s a podcast about how AI is changing media, news, and communication. I’m your host, Pete Pachal. I covered tech for a long time as a journalist, and now I talk with the media leaders, builders, and the creators, all trying to answer the question: how will we get information in the future? And how will that transform journalism in the business of media?
For most of the internet’s history, publishers had a pretty familiar bargain with platforms. They supplied information, and the platform helped people discover it. And at least some of that audience came back to the publisher. Well, AI is changing that bargain in a big way. Publisher content is now often summarized in an AI answer for users without even a single click resulting from that. Licensing deals sometimes compensate publishers for that use.
But payment, it’s really only part of the equation. The AI platform also owns the interface. That means it owns the behavioral data, often the advertising or subscription revenue, and, most importantly, the direct relationship with the person asking the question. All that has value. But just how do these licensing deals treat that value? Who’s getting the short end of the stick? And how is the market changing now that AI is everywhere and the right to scrape and summarize is constantly being repriced?
My guest this week has been tracking all of that more closely than almost anyone. Rob Kelly is the creator and host of Media & the Machine. He’s cataloged dozens of content licensing agreements and interviewed the executives, the deal makers, the publishers, and the data brokers trying to build this economy in real time. Rob also comes to the subject as an operator. He began his career writing about technology and interviewing leaders like Steve Jobs and Bill Gates. He later founded and ran media software companies. He’s also built early content syndication businesses, and he now uses AI extensively in Ongig and Daily Doc.
What I find especially interesting about Rob’s work is the tension inside of it. The number of licensing deals is growing, and the rights appear to be shifting from one-time training access towards live retrieval, or RAG, or attribution, if you want to call it that. But it’s not yet clear whether this becomes a durable new revenue stream for journalism in the media business, or is it just a transitional payment while the AI platforms take control of the discovery layer and the audience relationship?
I’m excited to get into all of that with Rob in just a minute. But first, quick note: if you’re listening on Apple or Spotify, please leave a five-star review and maybe a nice comment. And if you’re watching on YouTube, please like the video and subscribe to the channel. Those things really do help more people find the show. All right, let’s get started. Rob, welcome to the Media Copilot.
[2:58] Rob Kelly
How’s it going, Pete? I’m a longtime fan of yours. You have a great newsletter and podcast.
[3:02] Pete Pachal
Thank you, my friend. I appreciate that. So do you. But I’m really glad we’re finally chatting. I can’t believe it’s taken us this long to finally hang out on a podcast together.
[3:10] Rob Kelly
No. It’s fun. I’m psyched.
[3:13] Pete Pachal
Yeah. So for the folks who might not know you, tell us a little bit about your background. I gave kind of the CliffsNotes version just there, and the highlights of the famous CEOs you’ve interviewed. Give me a little more background on your relationship with both journalism and sort of the business side of media.
[3:32] Rob Kelly
Yeah. Well, I grew up on Wall Street. My dad taught me to love business. I was one of those nerdy little guys who, at 13, my dad dressed up in a Brooks Brothers suit and dragged me into the office as a kid, maybe 12, 13 years old. And he just taught me to love it like a game, kind of like folks like sports. I’d read the Wall Street Journal like I read the sports columns, and I love sports. And Dad told me I should write for the school newspaper, so I did.
He rarely gave me advice, by the way, so I took it. And that was a lot of fun. And then I started writing about tech, much like you. And I wrote for a company called CMP Media, one of the favorite companies I’ve ever run across, in Manhasset, New York, Long Island. They had great customers. So folks like Steve Jobs and Bill Gates would agree to sit down with me for interviews. And nothing to do with me. Just, I had a few hundred thousand people who are gonna write them a check. And you still remember the thrill of a byline, right? Seeing your own name.
[4:31] Pete Pachal
Oh yeah. First time you see your name in print, when it was print.
[4:36] Rob Kelly
Nothing like it. Yeah, nothing like it. And then I found that was like spectating, and I wanted to get in the game more. So I started running the content licensing business for CMP Media. And they had 16 magazines, a million articles. And my job, I was the only one who did it, was to sell that feed to LexisNexis, Bloomberg, Dow Jones News Retrieval. And it was a great business, highly profitable. I’m sure we’ll get into that later.
And that sort of whetted my appetite for the business side of things. I then founded or CEO’d three different startups. And my last startup, in early 2018, Ongig, I got lucky in that I had a smart co-founder, Kevin. And we started using Google’s AI back in 2018. This is four-plus years before ChatGPT. We were using machine learning, natural language, incredible stuff at the time, by the way. Still is, in fact.
And Google had us in twice to present to them. I’m so humbled by that, but we were one of their only customers too. So it was funny to have a big company like that. And Ongig was using their AI and paying them. So they had us in twice to talk to a 50-, 100-person team. Then I created this newsletter, Daily Doc, just for fun. And I started using ChatGPT because it launched around the same time. It 8x’d my productivity. So I was sold on AI.
And along came the New York Times suing OpenAI. I thought that was interesting. A few months later, the Wall Street Journal owner, News Corp, made a $250 million deal with OpenAI. And I thought, you love content like me, and media, I thought that was just the most unusual thing. Two of the top media companies in the world taking totally different strategies. So I launched Media & the Machine as a newsletter and then a podcast.
And we cover business, but at the end, also, I ask all my guests questions about sort of the future of AI and humanity, our families, our jobs. And I consider it a moral duty to be like a little guide out there. So it’s mostly business, but anyone’s gonna find the ending interesting because we’re talking about the future, and as you know, no one knows quite where this is leading.
[6:53] Pete Pachal
Nice. That’s great. And so I guess I’d love to get right into it in terms of, you mentioned those two big touch points in the AI content, well, licensing and legal stories of the past few years. They are very different in terms of the approach, but it’s all orbiting around the same thing. Can you tell me a little bit about those two deals? Basically, and those are all sort of in a context of a couple of years ago. But I guess I would wonder, what leads a media company to do one or the other? And then you tell me how the approach to AI may have evolved over the last couple of years since those deals were made or lawsuits were filed.
[7:41] Rob Kelly
Yeah, I think, I mean, the New York Times, that happened first, at the end of 2023. I’m staring at a spreadsheet, by the way, with 99 AI content licensing deals. I haven’t figured out how to share this, by the way, in a way that isn’t totally ripped off by AI companies.
[8:00] Pete Pachal
An AI guy like you? You haven’t vibe coded something with a lot of vibe coding? No. Give Claude Planner call. They’ll hook it up.
[8:11] Rob Kelly
Yeah, exactly. So what’s funny, though, is I started to log all these deals. Some went retroactively. So for instance, even though it was late 2023 that the New York Times sued OpenAI, shortly after ChatGPT launched, the New York Times had just done a deal with Google licensing their content, around a $100 million deal. It’s not talked about a lot, around 30 million bucks a year.
And I only use sources that are reputable, whether it’s Reuters or the companies announcing deals themselves. These numbers aren’t perfect. But the New York Times valued their data, their content. And so when OpenAI came along, they had their playbook. They were ready to protect that. They licensed their archives, just like I used to do for CMP Media, and they get paid for it. And along came OpenAI starting to use what the New York Times thought was their content. And so they were quick to sue.
[9:06] Pete Pachal
And so the $100 million, you’re right, it isn’t really talked about a lot. And honestly, the Google sort of deals and quasi-deals they’ve done have been a little different than the others in the space because they are so influential. But was that essentially to use the Times’ content in their AI products?
[9:27] Rob Kelly
You know, I’m guessing here only, and speculating, but I think it was such early days that they didn’t know exactly how things were going to pan out, but they wanted to be paid. So they had a relationship with Google for years. Obviously, Google sent a lot of traffic, and they’ve worked out issues over the years. But I imagine that the New York Times quickly got new terms in there, if they didn’t already at the start of the deal, to account for AI in some form, and renegotiated things, ’cause they’re the New York Times. They got some leverage.
[10:02] Pete Pachal
And certainly so did News Corp. So they had probably what was one of the most lucrative deals that I read about, at least. It’s all reporting, but apparently it was something like $250 million over five years, or whatever it was. So I found that interesting. And I think this sort of speaks to one of the insights that you’ve written about recently, which is that training data was sort of a big feature of those big deals circa 2024, it feels like.
And I’ve got a lot of questions about it. One of the main ones is, was that essentially a protective type thing? Because training data, to me, it feels like it’s already been done, right? It’s not like you can take the data out of the model once it’s built. And so I feel like it was sort of an after-the-fact licensing thing that would sort of come into this. And so you tell me whether my interpretation is sort of how this was treated at the time.
But I also feel like, as you’ve observed, training data is no longer the featured item in a lot of this stuff. And I’m kind of curious why. Do the publishers feel like that ship has sailed, especially considering some of the court decisions since that have sort of seemed to steer at least a little bit towards the fair use idea? Give me the whole background on training data and how it was treated back then, and why, as you’ve observed, it’s not as featured prominently in deals anymore.
[11:43] Rob Kelly
Yeah, and you’re right. So back in 2024, the vast majority of deals announced included training, including News Corp. And you’re right, five-year, $250 million deal. And by the way, just to put that in perspective, I know this because I ran content licensing for CMP Media. They’ve long since been sold, so I could share these numbers. I think we were up to about $2.5 million top line when I was licensing CMP’s content. Million articles, fair amount, two and a half million.
But that’s pure profit, except for my salary. We were turning on a feed, right? So it might not sound like as much, $50 million a year, but when it’s pure profit, that could be bigger profit than the majority of News Corp’s publications. So just a nice thing to keep in mind. This is not normal revenue. This is gravy, highly profitable. It’s a little different with AI, but the concept’s still the same.
And so you’re right. It was mostly training. I think it maps to how LLMs also evolved. In the beginning, they had to be trained. And then the second stage was you needed inference, as you well know, basically understanding how to answer questions then. And then finally you got into RAG. And sort of the deals have mapped that pretty well.
I still think training is hugely important, but what’s happened is that inference and RAG have caught up and been important as well. And training now is being used for different kinds of training. I’m sure we’ll cover this, but I have strong opinions on the fact that we have not run out of content and data to train on. Anyone who says we have is mistaken, in my opinion.
[13:30] Pete Pachal
Oh, what’s left? What are you basing it on? Are you thinking about multimedia or private data?
[13:36] Rob Kelly
Yeah, the way I look at it is that the whole world increasingly is recorded. And this has been happening over the last 20 years or so, as video recorders are now on our phones and everywhere. CCTV and all that stuff. So I look at it as endless, because do you remember in the old days when Google used to come out with something and say, X percent of the queries on Google this year, something like 35 percent, were new? And it always seemed really strange, right? You would think the queries would be like the Knicks score, I’m a Knicks fan, or what’s the news today?
But there’s so much new information that comes out every year. So I see that even more so with AI and training on content, in that every day we just are living new lives and we’re doing new things. More of it’s being recorded. Nearly 100 percent in some cases. And so, Madison Square Garden, they record the face of every person who walks in. Controversial. But picture that on steroids, or over in China, that’s done for just about everything.
And then you’ve got, of course, space, data going on in space, data inside your body, biology, data underwater, but even on land, the old-fashioned stuff, content and data. There’s still a tremendous amount of new data, and it’ll be needed all the time. And folks will still pay for training on that. It’s just the percentage of deals with training as the emphasis is lowering to around 40 to 50 percent.
[15:14] Pete Pachal
Yeah, but I guess my question was more like, do you pay for it because you don’t want to get sued? So that’s why you’re only going to pay people who have the resources to sue you, because you think that from a legal perspective, it’s defensible, right? It probably might be the most defensible part of AI scraping. And that kind of feels like maybe two things are a consequence of that. One, the number of people who you would make that deal with, who you’re scared of suing you, there’s a certain finite amount of them.
And then, well, I’ll let you go ahead and just answer, if you think maybe I’m just too much of a cynic about these things. But tell me what you think. And I know you’re more on the media side, but what is the AI company’s sort of attitude here, and why has it shifted?
[16:10] Rob Kelly
Well, I mean, I think you’re right about the training part. You can’t put the toothpaste back in the tube, as they say. And the models are well trained on what I would call the internet corpus currently, right? They grabbed all that. They grabbed it in many cases without either A, telling folks they grabbed it, or B, sharing where they grabbed it, which is, I think, by the way, a huge issue that is gonna come up more and more, that models are trained on data, they’re not being transparent about it. And that could lead to some pitfalls for AI companies.
When they’re using someone else’s data and a customer uses ChatGPT to create something, like at an enterprise level, like a new movie, where’s that footage coming from? So that’s my view. But I think the New York Times is probably a good case to talk about how, if you’re going to train on their content, they want to be paid. If you’re going to answer questions, inference on their content, they’re going to want to be paid. And if you’re going to show up today’s news stories on your LLM, they’re going to want to get paid. And I’ve got some strong opinions on also how they’re going to want to get paid. I think it’s going to change here, coming up, with the New York Times and Reddit.
[17:29] Pete Pachal
Yeah, well, tell me about that. I do feel like you’re 100 percent right, like inference and surfacing current news. I guess you actually set a few different kind of categories there. So maybe go a little deeper on what each of those means, and then what the compensation should be like, right? Because there’s a lot of different models, like pay-per-crawl, pay-per-use, which are similar, but obviously different in terms of when you get paid.
But then there’s other stuff that is associated with inference. We don’t have to go down through all the business models. But generally, with these deals in particular, are they favoring one type over the other? And then, for all of those, it feels like attribution is a key part of it. So has a standard emerged in terms of how that’s being tracked?
[18:35] Rob Kelly
Sponsored. Journalists, PR pros and communicators: the fall cohort of AI for Media starts October 13, six live Tuesday sessions with Pete Pachal plus two 1:1 coaching calls. Code AISEARCH500 takes $500 off the $1,500 price for anyone who found the course through AI search, a bigger discount than is offered anywhere else.
Well, you had a great interview with Jonathan Hunt, I believe, his name, over at HubSpot. Did I get the name right?
[18:39] Pete Pachal
Sure. Yes.
[18:42] Rob Kelly
And he sort of articulated it well, in my opinion, which is, I’m paraphrasing, but he basically said, we don’t really care how much you scrape our content. And they’ve got a media business, as you focused on in the interview. And all they care about is the business model they have on the back end to support it. So my belief is any content-driven company or media company just has to have a back-end business model to support this.
So let me give you an example: the most recent content licensing deal announced, a big one, is Yelp. And Yelp’s content, local data, this is with OpenAI. Yelp and OpenAI did a content licensing deal. And OpenAI gets to use all of Yelp’s local data and put it up there. It did not mention whether it’s used for training or not. But in Yelp’s case, what they’re getting is very interesting.
One thing they’re getting is a request for a quote. You know how you can be a service provider, like a plumber, on Yelp. And if you go on Yelp and look for a plumber, it says request quote. Well, that’s a very interesting new model, which is like a content to commerce, but it’s a licensing content to commerce model. And in a way, Yelp doesn’t really care that much on how much content they give up if they can get revenue from that request to quote.
And I think the New York Times is gonna do something similar. I think when they settle and do their licensing deal with OpenAI, you’re going to see it focused on generating traffic to their commerce, including subscriptions to the New York Times, subscriptions to all their digital stuff. You can buy it any way you want, their games and their recipes and so forth. So I think you’re gonna see more of that. And in that case, they don’t really care as much about being scraped if they’re getting the revenue on a usage base.
[20:41] Pete Pachal
So let me see if I’m hearing this correctly. So it’d be like, instead of just getting, your information has contributed to this answer and here’s a citation, they would also get, oh, and by the way, you can subscribe right here in the UI. There’s like a subscription button. Is that kind of the idea?
[21:01] Rob Kelly
Absolutely. Yeah. It’s kind of similar to affiliate ads and affiliate marketing, in a strange way. There’s no better phrase, although I’m starting to call it licensing content to commerce. Content to commerce is a known concept. So back to HubSpot, they’re happy. Jonathan’s happy to get a whole crapload of subscribers coming in from OpenAI, because that supports their business.
And I think the New York Times and Reddit is another example. One of the reasons they haven’t closed their renewal with Google, for instance, which is coming up here, is reportedly just that they’re holding out for more value of people clicking through to Reddit. They want to be the brand. That’s the other theme here. If you could be the brand, New York Times, Reddit, you’re in good shape. It’s a lot harder when you’re not the brand, of course, a known brand.
[21:56] Pete Pachal
Yeah, exactly. I know even USA Today is sort of musing, in the same way that Reddit was a couple of years ago, about, oh, maybe we’ll even nuke ourselves off Google entirely. I know the crawler situation is slightly different than it was before, but that just sort of seems like there’s a very few brands that could actually do that, that have the kind of leverage. And it is probably those three, the ones we’ve been talking about.
Do you put any stock in any of those kind of threats? I mean, I guess we’re not quite at Google Zero yet. We probably never will be at zero, but do you think those kind of threats will ever be real? Like, people wouldn’t mind, or not wouldn’t mind, but that the stakes of getting removed from Google would be less than whatever deal you think you’re not getting?
[22:51] Rob Kelly
Yeah, I mean, I think it’s gonna shift, but I also look at it as two scenarios. We’re talking about large publications, but if you and I were to form a startup tomorrow and suddenly, let’s say, we weren’t getting any traffic from Google, but we’re getting traffic from ChatGPT and Gemini and Claude, we’d be thrilled, right? So picture some 20-something launching a new business and all this free traffic’s coming on board. So I think it helps to take a beginner’s mind to this world.
The Google traffic obviously is going down for just about everyone I’ve talked to. And I know you’ve talked about this at great length, but taking a beginner’s mind and looking for what value can you get out of AI, and what’s your back end, to just run a good business, regardless of where the traffic comes from.
But I do think it’s gonna come from AI. That will get ironed out in the courts. You remember when ChatGPT launched, how there were no links, there were no citations, it was just pure answers? By the way, it was pretty, pretty great. But not fair, right?
[24:04] Pete Pachal
And then once they got links, half of them were made up.
[24:08] Rob Kelly
Okay. Yeah. And over time, it’s evolved to look more like, sort of, Perplexity did out of the gate, which is links and attribution. And I think all that will get ironed out, and there’ll be value coming from AI, and the good media companies with a back-end business model will prosper, but others are definitely gonna go out of business.
[24:30] Pete Pachal
Yeah, and it’s all, like you say, it’s all happening in the AI answer, and there’s so much focus on that. Your mention of the subscription button possible in sort of an AI answer is interesting. I’m definitely hearing, and there’s this stuff happening, that is suggesting that that real estate is now getting sort of rethought and renegotiated, right? So there’s the story from the other week about how TIME is now essentially trying to build ads for bots into their site, which is a very interesting approach, I find, just from a fundamental aspect, because it shifts the expectation of revenue away from bots and AI companies and to the advertiser. So the advertiser then essentially buys an ad specifically for the bots that gets scraped and passed along into the answer.
Which obviously invites a host of questions, and we’ll see how this whole thing plays out, as if it just actually becomes some kind of sustainable, or even any kind of, revenue. But I also see these things as a user too, and I’m like, oh, not that AI answers are clean, but relative to what a search engine results page looks like, they’re relatively clean today. And I feel like, oh, that, over the coming years, is going to be renegotiated.
And I even see it now, now that ChatGPT is advertising in the answers. They’re also linking out to brands more from the answers themselves. So there’s all these sort of things that are factoring into the user experience, the answer sort of getting reshaped. I guess my question is, how quickly do you see the AI answer becoming something more akin to the SERP result, with a lot of different things with different labels? And what are the chances that the publisher actually ends up getting a better deal than they got with Google?
[26:37] Rob Kelly
Yeah. I don’t know how it’ll net out in terms of versus Google in the end, but I will say that I believe that AI is going to increase the amount of queries and requests in the world, because we’re going to have these agents you’re talking about. I don’t know how many agents you have, but I have the equivalent of, I don’t know, five or 10. And I think the people really progressive right now that I’m talking to are well in excess of that.
And so you and I were talking about microphones, podcast microphones, before we got on. And if you wanted to get an upgrade to your microphone, and in the future you asked an agent to go look out for one, it’s going to go out and look at content sites. It’s going to run into an ad, as you pointed out, and that will influence the agent. The agent’s going to need new, current info, and someone will get paid for it. And I think it won’t be purely ChatGPT and the bunch. It’ll be the content and ad providers in that case.
[27:45] Pete Pachal
Yeah, yeah, exactly. And I think the idea of, I mean, it’s already a science, and a growing science, on sort of how do you influence these answers. And publishers have an interest in that too, because they’re competing in the information space. Of course, because their product is information, there’s often the sort of existential question attached, and then they just end up blocking.
So let’s talk about blocking for a minute, particularly for publishers. We’ve talked a lot about sort of the big publishers and the big players, but most folks, and certainly most of my listeners, are sort of in the mid- to small-tier range. And for them, there’s always the, well, I can’t get OpenAI on the phone, so I might as well just block. And this obviously applies to sort of everybody, which is to your point. The future is AI. If you’re blocking, you’re not going to be where information is now transacted.
So I’m oversimplifying here a bit, but I want to hear your take on what is a balanced approach here that helps preserve your authority, but also doesn’t just give away the store by just opening crawlers up to everything.
[29:02] Rob Kelly
Yeah, and I think this is where the squeeze sort of is. The large companies have the leverage of a legal team. So obviously the New York Times and others, big publishers and media companies, have that leverage to fight OpenAI and the bunch. And the small companies, as we talked about, taking a beginner’s mind, you might just not care. You just want to be discovered, right? You want your content and back-end products, hopefully you have them, to be found everywhere.
I think it’s that middle, the mid-size companies, that are feeling the biggest squeeze. And these are the ones that, I know you had Bill Gross on from ProRata, and TollBit and some others are trying to serve that world. And that’s the one where I believe the New York Times case is really gonna have an impact. Because I think they care about the money, but also what I’m hearing from their camp is they care about the principle. They actually really care about that. And I think they’re gonna show their leadership in this space by setting something of a playbook, is my hope.
And Reddit’s, by the way, doing something similar, even though they’re kind of different than a legacy media company, although they’ve been around now 20 years or something, haven’t they? So those mid-level companies, that has not been figured out. I mean, ProRata, I think Bill admitted, or shared, he wouldn’t hide this, but AI has not paid, as far as I know, a single one of ProRata’s publishers. And so they got some things to figure out there. And it’s not ProRata’s fault. I mean, it just hasn’t worked itself out.
But there are some interesting new protocols coming up, as you well know. I find the RSL, Really Simple Licensing protocol, that one is one to watch, because they have about 40 percent of the internet’s content companies behind them. So that’d be a collective, so that they can go to OpenAI and say, we represent Reddit, which they do, and they represent People Inc. and a bunch of others, and it’s time to sit at the table.
And the good news about them is they represent the big companies, Reddit and People Inc. and so forth, but anyone can sign up. You and I can sign up for our podcasts, and we can write into the code that, hey, we want to get paid if you scrape our content, or we only want traffic, we don’t want to get paid. And you actually write a contract within the code. I think you’re gonna see some protocols like that become standard.
[31:50] Pete Pachal
Yeah, it definitely does feel like, slowly but surely, the media side of things is figuring out collective action, even though it is not sort of inherent to their DNA. You mentioned RSL. There’s Spur, that started in the UK, that’s trying to sort of figure out, at the very least, what the plumbing should look like and how to correctly attribute things. And there’s just a lot of collective activity that didn’t exist a couple of years ago. It took ’em long enough.
[32:22] Rob Kelly
And by the way, just in a strange twist, we’ve always been so far ahead, the US, ahead of the UK and the rest of the world in tech. Things have really tightened up in terms of AI. And there are some ways in which the EU and some of their new laws and Spur and so forth might take off fast over there, and we might actually follow suit because we see a new model. Just might be an interesting twist to look at.
[32:52] Pete Pachal
Yeah, and I know I’m interested in seeing how Google ends up allowing folks to sort of split off their AI Overviews. Because right now, I believe you can opt out of Gemini through Google-Extended, but you can’t opt out of Overviews. And I feel like it’s funny, Google claims, a lot of folks claim, we’re never gonna put ads in our AI experience or our chatbot in the way OpenAI does, but yet they have all these AI experiences throughout the rest of their product that is ad-supported anyway. Just find that a bit of an irony.
[33:29] Rob Kelly
And by the way, how cool would it be if Google took their page out of the YouTube creator monetization model, which they pay 55 percent? You probably know, you’ve got a YouTube channel: 55 percent of the ad revenue over to the creators. Huge opportunity for Google to do this with all creators. They’ve already got the model, they’ve got the tech. And I could also see Apple doing something like this. They did it with iTunes. And they took Napster, an illegal sort of marketplace of music content, and they turned it into something more like HBO Max, or just a subscription model.
And I think there’s some opportunities for some companies to make some big moves. And the AI companies are not doing so well in terms of sort of favorability out there. And they might face some challenges here. Some friends of mine are saying the pitchforks might be coming for the AI models and their CEOs, with folks being so upset about what they’re doing, young folks especially. Aren’t you hearing this from younger folks in your life?
I mean, anyone under 18 or 20, I would say, for me, it’s about two out of three, or maybe even three out of four, totally down on AI. I’m an AI optimist, so it’s hard for me to hear that, but young folks can make some noise.
[34:55] Pete Pachal
Yeah, there’s 100 percent a backlash, for a bunch of reasons. And the two main ones that I feel like are most influential are, the teachers are mostly still skeptical of it, and the message, AI is cheating, goes out to most kids in K-12 and, to a large extent, folks in college. Some folks are AI-forward and are sort of more acknowledging that this is the future, and sort of teaching kids how to use it, but that’s rarer.
And the other part is sort of all the stuff going on in society now with regard to just activism, whether it’s anti-data center, anti-AI, or just sort of this general unrest that’s kind of almost class warfare, which involves suspicion of big tech, et cetera. There’s all this kind of stuff that circulated around that. And I think that has an outsized influence on the younger generation. So, yeah.
[35:50] Rob Kelly
And their jobs. Yeah, they’re real worried about their jobs.
[35:53] Pete Pachal
Well, this is the other thing. Again, I don’t want to get this into too general AI commentary, but the whole idea that the AI marketing is always about doom and gloom, which is kind of, it’s come back around. And there’s very little marketing of AI like, hey, this is going to do great things for you. But I also know you and I, I guess I think of myself as an optimist too. And we’re seeing great results just in our personal work.
Like, you’ve said you’ve kind of 8x’d your productivity, which I’d actually like to double-click on there a bit, because that’s interestingly precise. And I’m glad you didn’t say 10x. How did you come to that?
[36:28] Rob Kelly
No, I can tell you it was 8.7x. And it’s because I measured it in a spreadsheet. I measured in a spreadsheet with Daily Doc, as I was writing. It was purely word count. I can’t spam, it’s not in me. So I write the same quality content, and I measured my word count from blogs pre-Daily Doc that I did for my SaaS company versus Daily Doc. And I just looked at word count, and it was 8.6x the volume in the same periods of time.
So it meant I could do it faster, I could do more, whatever you wanna measure it. But that’s what got me sold on AI. I was like, wow, that’s my own personal thing. And I’m not that good. There’s others who are doing it 100x.
[37:14] Pete Pachal
Yeah, I certainly made more liberal use of AI as time’s gone on. But this year, with the rise of co-working apps and agents, it’s really night and day. It’s been difficult to measure, other than I’ve been able to do things I could never have done on my own, which is to say a lot of the coding and maintenance of software and sites and that sort of thing that you need a whole team to do. Well, now I do have a team. It’s just bots and agents that run it for me, which is great.
So as we’re wrapping up here, I want to just get, honestly, your view of the future. And I’m speaking specifically about licensing business models, the future for publishers and media. What do you see? And I wouldn’t mind if you sort of game this out in sort of a variation of my final question, which is, how do you game that out where media loses, and how do you game it out where media wins? And you can define those terms however you like.
But I feel like most people in media feel like we’re on the losing path. Give us some thoughts on why that might not be the case, but what would happen for both of those sort of endpoints.
[38:38] Rob Kelly
Yeah, I think, well, I’ll take the argument for, and I love that. I love Peter Thiel’s contrarian philosophy, by the way. Like ’em or hate ’em, or somewhere in between, he always takes the other side. And I think that’s a useful exercise for us as humans, beyond just AI talk: humanity. So making the case that media could really take a big hit would be that AI companies win most cases, including the fair use. And they argue it’s transformative, so they can train on any data, content, they want.
And we’re just slow to regulate things like compensation for any media company for using any content. I think the middle folks, the mid-tier companies, media companies, content-driven companies, a lot of them could take a big hit. And if they don’t have good business models, a lot are gonna lose. Sort of like the small mom-and-pops back in the Amazon days. But I would call it the medium-sized mom-and-pops, not just small creators, because they can do quite well. So I think it’s gonna be a really tough time.
But also, I’m a little bit of a Darwinian sort of thinker about this, which is there are a lot of companies that weren’t healthy enough. All good things come to an end. That’s gonna happen. That’s us evolving and getting better. So I believe that.
On the flip side, I think the media companies have enormous clout and leverage. So, for instance, I believe that OpenAI will settle with the New York Times. They’re gonna want to settle before their IPO. Sam’s gonna want to beat Dario at Anthropic in any way he can. And I think OpenAI will close a deal with the New York Times, and it’ll be transformative to the industry. And then they’ll have a blockbuster IPO. And there’s plenty of money to share.
[40:55] Pete Pachal
I was just thinking that’s gonna be a big check. Especially if the New York Times, as you say, is really trying to stand on principle here. But everybody has a price.
[41:06] Rob Kelly
And some of the thinkers here, including Sam Altman at OpenAI, have talked about this idea of sharing all the wealth, not just universal basic income, but there’s a new thing called, I forget the name for it, but it’s like sharing just the capital, sharing the market value of these AI companies. It’s possible we see some bailouts, and things might change, and it’s just a value distribution. So as well as OpenAI and Anthropic all do, maybe that money gets flowed back into our pockets, whether it’s a company or an individual.
So again, I suffer from optimism bias. I think things will work out in humanity, because we’re just an amazing species, but they’re gonna change, that’s for sure.
[41:58] Pete Pachal
Nice. Good stuff, Rob. We’ll leave it there. Thanks for dropping by, sharing your thoughts and all your analysis on all the licensing stuff that’s happened in AI in the last few years.
[42:10] Rob Kelly
Hey, Pete, keep up the awesome work.
[42:12] Pete Pachal
Appreciate it, man.
[42:14] Pete Pachal
Thanks for listening to my conversation with Rob Kelly, creator and host of Media & the Machine. The Media Copilot is a podcast. Podcasts are great things to subscribe or follow on whatever platform you’re listening to us on. If you’re on Spotify or Apple, please leave a five-star review and maybe a nice comment. And if you’re on YouTube, please like the video and subscribe to the channel. All of those things really do help the show because they put the podcast in front of more people.
Don’t forget to subscribe to the Media Copilot newsletter as well. Thanks for being here. Be seeing you in the future.
