Publishers are currently salivating over a phantom. They hear the number "100 billion" and imagine a rain of licensing checks falling from the clouds of Silicon Valley. They think they are finally in the driver’s seat. They are wrong. They are effectively rearranging the deck chairs on the Titanic while hoping the iceberg pays them for the privilege of the collision.
The current industry consensus—championed by groups like RSL—is that AI companies owe a massive, ongoing debt to news organizations for training data. The logic is simple: "You used our facts, our reporting, and our IP to build your trillion-dollar models. Now, pay up." It sounds fair. It sounds like justice. In reality, it is a desperate attempt to apply 20th-century copyright logic to a 21st-century intelligence engine.
The Licensing Trap is a Race to Zero
The idea that news is a $100 billion opportunity for publishers in the AI age assumes that the value of information remains high as the cost of synthesizing it hits zero. This is a fundamental misunderstanding of supply and demand.
Licensing deals are not a new revenue stream; they are a settlement. When a publisher signs a deal with an LLM provider, they are essentially selling their long-term relevance for a short-term cash infusion. These deals are structurally designed to favor the buyer. Why? Because the LLM doesn't need your specific article. It needs the information contained within it.
Information is a commodity. Facts cannot be copyrighted. If a local news outlet reports on a fire and three other outlets rewrite that report, the "fact" of the fire is now public domain in the eyes of an LLM's training weights. The model learns the event, not the prose. By the time the licensing check clears, the model has already extracted the only thing it actually wanted: the data points.
Your Archives are Depreciating Assets
I have seen media executives treat their archives like a gold mine. "We have fifty years of reporting!" they boast.
Here is the cold truth: Most of your archive is worthless to a frontier model. AI companies care about high-quality, reasoning-heavy data and real-time updates. Yesterday’s news is just noise in a sea of petabytes. The "100 billion" figure being thrown around is a collective hallucination based on the total market cap of AI companies, not the actual utility of a Tuesday morning op-ed from 2014.
The "landlord" model of digital media is dead. For two decades, publishers lived off the rent of attention. They sat between the reader and the information. AI removes the middleman. If a user asks a bot for the top three takeaways from a complex geopolitical event, and the bot provides them accurately, the user never visits the publisher's site. A licensing fee of a few pennies per query doesn't replace the loss of the entire relationship with the audience.
The Fallacy of the "AI Subsidy"
Critics argue that without news, AI models will "hallucinate" or lack "grounding." This is the leverage publishers think they have. It’s a weak hand.
Synthetic data and specialized "agentic" search are already filling the gaps. Companies like Perplexity or OpenAI’s SearchGPT don't need to index every word of your 3,000-word feature. They need the juice. They need the signal. And they are getting very good at finding that signal in places that don't charge a premium.
When you demand a $100 billion payout, you are telling the tech giants that your data is a luxury good. Their response won't be to pay forever; it will be to find a cheaper substitute. We are already seeing "data contamination" fixes where models are trained to prioritize certain reliable sources over others, but that list of "reliable" sources is shrinking, not growing. If you aren't in the top 0.1% of global brands (think The New York Times or Reuters), you have zero leverage.
The Quality Paradox
The irony of the licensing movement is that it incentivizes the very thing that will kill news: volume over value.
To maximize licensing revenue, publishers are tempted to churn out more content, more frequently, to "feed the beast." But the beast is already full. LLMs are already suffering from "model collapse" caused by eating too much AI-generated junk on the open web. If you start producing content specifically to be "scraped" and licensed, you cease being a news organization and start being a data farm.
Data farms are low-margin businesses.
Why the "Fair Use" Fight is a Distraction
Publishers are pouring millions into legal fees to fight over "Fair Use." This is a tactical error. Even if the courts rule in favor of publishers, the victory will be pyrrhic.
Imagine a scenario where every AI company is forced to pay for every snippet. The result isn't a thriving media ecosystem. The result is a consolidated market where only the biggest tech firms can afford to operate AI, and only the biggest publishers get paid. Small, independent, and local journalism will be locked out of the "revenue share" because the administrative cost of tracking their microscopic contribution to a model's weights exceeds the value of the data itself.
Stop Asking for a Cut and Start Building a Moat
The $100 billion "opportunity" is a distraction from the real work: building a business that doesn't rely on being a parasitic attachment to a search engine or an LLM.
If your business model depends on a third party scraping your site and then giving you a "fair" share of the profit, you don't have a business. You have a hobby that is being tolerated by a tech giant.
The only way to survive is to create "non-scrapeable" value. This isn't about paywalls; it's about community, live events, proprietary data tools, and personality-driven reporting that an AI cannot replicate because the value is in the identity of the creator, not just the information shared.
The industry is obsessed with "compensation" for the past. They should be obsessed with "utility" for the future.
The Hard Truth About Traffic
We need to stop lying about "referral traffic." For years, Facebook and Google fed the industry a line about how they were "partners" in sending traffic to news sites. Then they turned off the faucet.
AI companies will do the same, but faster. They won't send you traffic because the goal of the interface is to provide the answer in situ. If you are waiting for a $100 billion check to save your newsroom, you are waiting for a ghost.
The money isn't coming to save the journalism we have. It’s coming to buy the scraps of the journalism we’re losing.
The real opportunity isn't in licensing. It's in the realization that the old web—the one where you wrote a story, Google indexed it, and you sold an ad—is over.
You cannot litigate your way back to 2012. You cannot tax your way to innovation. The $100 billion isn't an "opportunity" for publishers; it's the price Silicon Valley is willing to pay to clear the path for your replacement.
Stop signing the settlement papers and start building something that people will actually go out of their way to find when the bots stop telling them where to look.
