The Economics of an AI Future Doesn’t Work for Publishers
Fifty million dollars per year. That’s what News Corp got from OpenAI to license its content. It’s the kind of offer that would have any publisher envious—especially when the profit margins approach 100%. But the fundamental problem is none of this licensing revenue can ever offset the massive drop in revenue due to a loss of audience.
And it’s not just News Corp. Many others have done deals, including:
- Dotdash Meredith is getting approximately $16.3 million per year
- Axel Springer is generating in the eight figures per year
- The Atlantic and Vox Media both have deals for an undisclosed amount of money
The money’s good, especially when you consider that it’s all licensing revenue, so the margin percentages are in the high 90s. When you’re an executive that needs to deliver on a certain EBITDA target, it’s hard to look at these deals and say no to them.
My opinions about these sorts of deals have evolved over time. At first, I was against them entirely because of the disintermediation of the audience. Then, I thought about it and said, “I can’t control where the audience wants to go, but if there are multiple platforms, maybe publishers can get licensing revenue from each of them.”
I remember chatting with a person at Condé Nast a couple of years ago and they said they were having conversations with a number of firms—Apple included. If you can stack eight-figure deals from a number of firms, the money really starts to add up. And again, the margin is good because what’d it cost you to deliver on that licensing agreement? Basically nothing.
Fair use or fair theft?
A big reason why it has made sense to take these deals is because many of these AI companies do not respect content creators and the spirit of fair use in the context of copyright. If the content is going to be stolen anyway, shouldn’t you get paid something for it?
A number of weeks ago, I spoke with Paul Gerbino, President at Creative Licensing International about this, and a couple of folks from his team. One thing Gerbino said has stuck with me:
When did fair use become a right, not an affirmative defense? There’s this attitude by Big Tech that they have this God-given right, they’re endowed by their creator to take all this stuff [content] to do the things that they want to do. And that’s part of the challenge.
I won’t bore you with all of the examples of tech companies stealing first and calling their lawyers second, but in no world would anyone believe that ripping every single piece of content from the internet is fair use. The scale of it is incredible.
And so, I understand why these licensing deals happen. With a steal first, ask forgiveness later policy, media companies are hoping to get anything they can from tech companies. But by accepting that money, you’re effectively forgiving the theft of your content and accepting far less than the content is worth.
The problem is that no single company can fix this. Why should I expect Dotdash Meredith, News Corp or Axel Springer to fix what is, fundamentally, a legal and internet economics problem? There are attempts to fix things on the legal side. The News/Media Alliance is suing Cohere, which, if they win, would be a good step for publishers.
But from an economics perspective, things are much worse because it blows up how we make money on the internet. Consider this…
A drop in AI traffic
A user goes to ChatGPT and asks a question. They get the answer they want. There are sources listed and little grey boxes that tell you where you can get more information. But I got the answer already, so why do I need to click over? This hypothetical is what a ton of users are doing every day. Tollbit’s Q4 2024 State of Bots report found that AI chat bots “on average drive referral traffic at a rate that is 96% lower than traditional Google search.”
If 96% of people that would have landed on your site don’t, you’re out of business. No amount of licensing revenue can offset that because the licensing revenue doesn’t take into consideration the true LTV of a reader. That’s why the economics are broken.
When someone lands on your site, you monetize them with ads. Very low revenue, but it’s something. You could get them to subscribe to a newsletter. In that newsletter, you run more ads. That’s more revenue. They click over to your site a number of times, hit the paywall and convert. That’s more revenue. Finally, you host an annual event and they decide to buy a ticket to said event. Even more revenue.
Now, not everyone is going to do everything. But with any good funnel, some percentage of these folks will. The LTV of those readers is far higher than any amount of money that you’d get from a licensing deal. Let’s do some math using Dotdash Meredith as the example.
Math madness
In 2024, DDM had 10.6 billion sessions across its owned & operated sites. In that same year, it generated approximately $1 billion. If we do basic arithmetic, that means it earns $0.094 per session. Now, we’re going to make a number of assumptions:
- 50% of DDM’s total sessions come from Google. That’s 5.3 billion.
- Traditional Google will lose 50% marketshare to chat-based solutions.
If the second assumption occurs and chat bots drive 96% less traffic than Google, DDM would see traffic go to about 8 billion sessions. That’s a loss of 2.6 billion sessions (this math does assume the 4% that does click over from a chat bot). If DDM monetizes at the same rate, it would have generated $757.26 million in revenue—nearly a quarter million off from what it actually generated.
These are aggressive assumptions, but they’re necessary. DDM generated $106.9 million in operating income in 2024. DDM goes from an incredibly profitable business to a money-losing one if chat-based solutions become the default.
So, what’s the answer? I’ll let DDM’s CEO, Neil Vogel, answer that. At a Press Gazette event, he said:
The basic tenets of media are you have to have great brands, you have to have great audiences, you have to treat them with the utmost respect and love. You will then figure out how to make money if you have these things. If you don’t have these things, and you try to do this, you will go to zero.
…
You have to understand that brands are like children. They’re all totally different. And what we do with Food + Wine, what we do with Verywell and what we do with People do not resemble each other in any way.
…
If you make yourself essential, you’ll be fine.
He’s right. The problem is that it requires a complete rethink of how operators have built on the internet. You need to be essential. You also need to stop thinking traffic is a passive thing and increasingly invest in other more active channels (see: paid growth). I don’t know if Vogel would agree with this, but being an exclusively ad-based publisher will become increasingly difficult because you’ll need to generate more money per person that does hit your site—aka increase the LTV.
The economics in a AI chat-based world don’t work for the free web. Unless something changes, there’s a strong argument for the internet becoming a bunch of walled gardens. In many respects, we are going back to a pre-internet era where people had to seek out individual brands to get their information rather than relying on aggregators. If you stay still, you die. You must always evolve.
