As the news media industry moves from a traffic era — with well-defined value exchange and controls — into the generative AI era, value, incentives, and safeguards are far less clear.

That is forcing publishers to change their mindset and actions, explained Madhav Chinnappa, senior executive consultant and researcher at the Reuters Institute, during this week’s members-only Webinar.

“I think we need to acknowledge we’re in a very different era,” he said. “We need to think differently, and we need to reinvent the underpinnings of this so that we have a sustainable information ecosystem.”

In an hour-long, far-reaching conversation with Jodie Hopperton, lead of the INMA Product & Tech Initiative, Chinnappa discussed the evolving relationship between AI foundation model companies and the news media industry, as well as the challenges news publishers face in licensing content to large tech companies. 

A pivotal shift

Chinnappa, whose extensive background includes roles with BBC, Google, and Human Native AI, noted the structural shift the industry is experiencing is more profound than the arrival of the Internet itself.

In the “era of traffic,” he explained, publishers allowed search engines to crawl their content, and in return they received traffic. There were controls — such as robots.txt — and incentives aligned with producing legitimate content.

But that has all changed: “In the era of gen AI, I don’t know what the value exchange is, I don’t know what the controls are, and I worry what it does to legitimate content creation,” he said.

This new era touches every part of the news business simultaneously — from reporting and production to distribution, monetisation, and user experience. Generative AI is already reshaping how audiences consume information, shifting behaviour from search engines to answer engines.

“I suspect in a year or two, if you explain to an 8- or 9-year-old kid that you were typing things into a search box with some words and then you’re getting a list of links and then you clicked on them to get what you wanted, they’d kind of look at me like my kids look at me when I explain what a fax machine is,” he joked.

One of the most sobering insights Chinnappa shared is that news content, despite its civic importance, is not strategically valuable to foundation model companies.

“News content is actually not that important … from a licensing perspective,” he said.

“That’s kind of hard to come to terms with because — and I say this with love in my heart as a news person — that the news industry always consider themselves quite important. And they often talk about pillar of democracy and all that, which I completely agree with. But when you get down to how this works, what they’re trying to build, news is just not that important.”

His experience at Human Native AI, a startup later acquired by Cloudflare, reinforced this. News is abundant, interchangeable, and widely available on the open Web — making it a poor fit for AI training data marketplaces. Foundation models, he explained, prioritise content that is non-commodity, scarce, structured, and monetisable. Academic, scientific, medical, legal, and technical content fits this profile.

News does not.

This leads to a deeper concern: If exclusivity evaporates the moment a story is published and becomes instantly ingestible by AI systems, what incentives remain for original reporting?

“There’s actually no such thing as exclusive anymore,” he warned. “And if there’s no such thing as an exclusive, why would you go and do original content? Who’s going to do all of that reporting?”

An ecosystem at risk

Hopperton pressed Chinnappa on the societal implications of losing professional, fact‑checked content. If publishers cannot sustain their businesses, foundation models will be trained on lower‑quality, unverified, or politically motivated material.

Chinnappa agreed the risk is real — but not yet a priority for AI companies. He described the dynamic with a vivid analogy: foundation model companies are “having this kind of world war against each other up here,” while publishers warn from below that “the water system is getting polluted.”

Eventually, he argued, AI companies will have to face the consequences of these polluted, degraded information ecosystems. But for now, competitive pressure dominates their attention.

“This is where I think we need to act collectively,” he said. “Before, in the era of traffic, we didn’t need to act collectively. We didn’t think collectively. But now we need to think much more collectively because we’re in this different era.” 

Collective action as the “least worst option”

A major theme of the day was whether news publishers should pursue collective action. Hopperton polled the audience on how likely they were to consider acting collectively. While 29% said it wouldn’t happen, 33% were open to it — a higher number than Chinnappa anticipated.

Collective action is not only viable but necessary, Chinnappa said: “I consider it what I call the least worst option,” he said, explaining that blocking alone is insufficient. “If you’ve stopped at blocking, then you’re not going to get much out of this.”

But blocking, combined with collective action and a product‑oriented licensing model, could give publishers the leverage they currently lack.

“I’m trying to set up a situation where the news ecosystem leverages what few points it has for the benefit of everybody in the medium- to long-term,” he explained. “So that’s why I start with the collaboration. I think you then block, and then what you’re licensing is access.”

Hopperton asked Chinnappa to go deeper into what collective action is — and how publishers and news media organisations can work together to move forward.

“The way that my small brain looks at it is it’s really about four questions. It’s who, what, how, and how much,” he said before breaking each one down:

1. Who: Eligibility guidelines at the country level, allowing any qualifying publisher to participate. “In some countries, like in France and in Canada, there’s a list, but you have eligibility guidelines,” he explained. “So anybody who wants to be part of it can be part of it.”

2. What: A shared licensing proposition focused on access — particularly structured, real‑time API access — rather than raw content.
3. How: Coordinated blocking of AI crawlers, followed by offering structured access as the licensable product.
4. How much: The hardest question, requiring experimentation and negotiation.

The conversation provided a new perspective on collective action, Hopperton said:

“I think we have a habit of thinking sometimes that collective action is about regulatory and I don’t think that’s what we’re talking about here. I think it’s really about how can we speak as an industry because that will probably make it easier for the tech companies as well.”  

As the Webinar closed, Chinnappa offered a final challenge to publishers: “I think we as an industry need to build the future that we want rather than accept the future that we fear. I think it’s really important. I think so many people feel almost demotivated, powerless or whatever, and I get it. I truly do.

“But I think you’ve got to try. You’ve got to try to build something.”

. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .