This week’s Subscribed Weekly comes courtesy of Amy Konary, Senior Vice President at the Subscribed Institute
Meet journalism’s dream reader.
It is insatiably curious. It can read thousands of articles before breakfast, compare sources, check facts, monitor breaking news around the clock, and come back tomorrow hungry for more.
There’s one significant problem: it may never visit your website, see an ad, or buy a subscription.
Because it isn’t human.
That may sound like bad news for publishers, but I’m increasingly convinced it could be the beginning of a much bigger market for journalism, if publishers can turn that new demand into durable revenue.
Which raises the harder question: What happens when your most voracious reader never becomes your customer?
For most of the digital media age, consumption and monetization have been connected. A person finds an article, visits a publisher’s site, and creates an opportunity to earn revenue through advertising, subscriptions, or another direct relationship.
AI can separate those events. An agent might monitor hundreds or thousands of news sources, compare reporting as a story develops, verify new information, and synthesize what matters into one answer. All of that consumption could happen without generating a page view, an ad impression, or a new subscription.
Media companies could face a strange new condition: journalism is becoming more valuable even as it becomes harder to monetize.
Generative AI makes synthesis abundant, but producing new, verified knowledge remains costly…and human. Someone still has to attend the city council meeting, interview the CEO, call the second source, and verify the number. If something happens in Boise and only one or two local organizations report it firsthand, that information may become disproportionately valuable because few substitutes exist.
A recent INMA report by Pradeep Gairola of The Hindu calls this the Original News Production Layer. As summaries, rewrites, and synthetic articles become abundant, new knowledge about the real world becomes comparatively scarce.
The companies building these systems have worried about the same dynamic. Newly unsealed court documents reported by The New York Times show that people at Microsoft and OpenAI worried AI could weaken the news ecosystem their models rely on. One Microsoft researcher described large AI models as “a product that destroys its supply chain.”
That line gets to the heart of the problem. AI can make original reporting more valuable while weakening the economics required to produce the next piece of it. Eventually, that becomes a supply problem for AI itself.
That is also one reason I’m optimistic. AI companies have an economic interest in preserving access to high-quality, original information. If systems become dependent on journalism, sustaining that supply matters to both sides of the market.
Scarcity creates economic value. Dependency creates negotiating leverage. Neither automatically creates revenue.
The missing economic layer
Plenty of people are thinking about this problem. What makes me more optimistic is seeing companies start working on the pieces a solution would actually require: identifiable machine buyers, negotiated access terms, and mechanisms for measuring paid use.
Matthew Prince, CEO of Cloudflare, is also a local newspaper owner who cares deeply about access to quality reporting. Parag Agrawal, founder of Parallel, is building infrastructure for an internet increasingly consumed by AI agents.
They are working on different parts of the same problem.
Cloudflare is helping publishers see and control machine access to their content. Cloudflare’s Pay Per Use beta creates a market where AI buyers can propose payment for a defined use of publisher content, with Cloudflare facilitating the transaction. Parallel is working on ways to measure how much individual sources contribute to an agent’s output. Both are trying to make machine consumption visible and measurable enough to value and monetize.
These markets are early, but something important has changed. And the human reader still matters enormously. Publishers spent years building direct relationships with those readers, and those relationships remain strategically valuable. The goal is to add the machine customer without surrendering the human one.
A plausible future is one in which agents dramatically increase demand for trustworthy information, original reporting becomes more valuable, and publishers develop recurring commercial relationships with the systems that depend on it. Journalism could gain a larger economic base than it has today.
There is also a future in which AI consumes far more journalism while the economics flowing back to the people producing it remain too small, too unpredictable, or too concentrated among a handful of technology platforms to support the work.
The next billion readers may not be human. But the reason we need to get the economics right absolutely is.
