‘We don’t use AI anymore – we live in it’: Three years after ChatGPT’s debut, publishers confront an existential crossroads

"It feels like each year, the stakes are getting higher," said Ezra Eeman, Strategy and Innovation Director at NPO, the Dutch Public Broadcaster, and lead of WAN-IFRA's AI in Mediainitiative.

Eeman gave an overview of the current AI landscape At our recent Paris AI Forum, taking stock of near-future risks, opportunities, and pressing decisions facing news publishers.

Things in this space are moving dizzyingly fast, as users are adopting AI tools such as ChatGPT, Sora and Meta AI at an unprecedented rate. Meanwhile, AI search is predicted to overtake traditional search as a web traffic driver in 2028, and AI-generated content is likely to soon outnumber human-generated content online.

In short, the landscape is shifting rapidly under news publishers' feet.

"We're on a roller coaster. AI has taken us on a very rapid ride that simultaneously promises to solve everything, but also threatens to disrupt everything," Eeman said.

• Ezra Eeman will be speaking at our Newsroom Summit, which will take place on 18-19 November 2025 in Copenhagen, Denmark.

'For consumers, this isn't hype'

Amid all the talk of AI hype and potential bubble, it is helpful to consider how AI tools are actually being used, with ChatGPT leading the way. Eeman pointed out that OpenAI's chatbot now processes an astonishing 2.5 billion prompts a day.

Much of this usage comes from younger generations. In the Netherlands, for instance, 86 percent of 15- to 20-year-olds actively use AI, and ChatGPT's mobile users spend an average of 16 minutes per day on the app.

"So for consumers, this isn't hype," Eeman said.

"AI is becoming as common as social media, certainly for younger generations. It's quickly becoming a new habit to go to chatbots to ask for anything - things for school, things for work, inspiration, anything."

However, some major pain points prove to be difficult to solve, such as the fact that AI systems "still struggle with news," Eeman said. A recent EBU study foundthat 45 percent of AI answers about news had errors, such as issues with sourcing, hallucinations and outdated information.

EBU's study concluded that almost half of analysed AI responses included flawed answers, while a third showed serious sourcing problems.

AI systems misrepresenting news content is still "a significant gap, and it hasn't been closed much since a year ago," Eeman said.

Moreover, the threat of "Google Zero," or a collapse in traditional search traffic, remains very real. Figures vary, but some publishers have reported declines of up to 40 percent since Google added AI Overviews.

More data is emerging from early adopters and publishers that have partnered with AI platforms and are sharing their findings, and "it's not all negative news," said Eeman.

Schibsted, which partnered with OpenAI, said its brands are receiving millions of impressions from ChatGPT. Opinion pieces in particular are performing much better in this environment, with a click-through rate of 5 percent compared to just over 1 percent for news, sports and business content.

However, Eeman warned that a "winner takes all" dynamic is starting to form in this area. Larger brands are surfacing more regularly on AI platforms, while smaller publishers have less visibility in these environments.

From 'AI in Media" to 'Media in AI'

Beyond these developments, Eeman also estimated that more substantial, even if less obvious, changes are happening.

"Beneath the surface, I think an even bigger thing is brewing," he said.

"The environment itself is completely changing, and we might easily miss it."

He referred to this ongoing shift as a move from "AI in Media" to "Media in AI."

"In this new model, everything is flipped. AI is no longer just an add-on. It becomes a primary interface through which people experience media."

Chatbots represent only the first step in this direction, as "our content is becoming an ingredient within an AI answer, rather than a destination itself."

Beyond chatbots, AI systems are being expanded to become not just content consumption platforms but complete operating systems in their own right. The recent introduction of apps within ChatGPTis a prominent example of this, but it's just one instance of AI companies building comprehensive ecosystems around their AI solutions.

"These aren't tools anymore. These are complete environments. And if you're not figuring out how you're visible and relevant in these ecosystems, you might as well not exist," Eeman said.

From a user's perspective, this means that content consumption is likely to become more ambient and personalised, with AI systems tailoring content to meet users' needs.

"In this scenario, users never visit a website or an app, because everything is orchestrated by an assistant in the operating system," Eeman said.

While strong destinations are likely to remain relevant, he said publishers must consider the new dynamics of AI systems that surface content based on users' needs, mood and context.

Next step: agents taking over?

While most of us are still adjusting to the current wave of AI, "the next one is already building up," warned Eeman.

With the rise of autonomous AI agents, "we're on the brink of an AI that not only finds things, but also acts on them."

The push to include AI features into web browsers and the launch of new browsers by AI companies is part of this trend. The aim is to become intermediaries for online consumption in order to collect as much user data as possible, and eventually act on users' behalf.

"The playbook is the same for all of them: to own the user's entire journey and the workflow, so you never have to leave the browser to get things done."

In the agentic future, AI tools will interact with websites in the same way as people do: visiting sites, collecting information, filling out forms and doing our shopping. Consequently, media companies need to find a balance between optimising their content for human readers and AI assistants.

One possible framework for making sense of this is the funnel model. In this approach, capturing the attention of AI tools with topic authority and well-structured content forms part of publishers' top-of-the-funnel strategy.

Second, building trust and creating distinctive content will help to deepen the relationship with audiences and encourage them to visit websites directly.

The final step is to transform that trust into revenue by protecting the most valuable content (in a format that cannot be scraped) and helping to convert loyal users into paying customers.

Publishers have largely jumped in to experiment with new AI tools, and crucial learnings are emerging from these trials. But media companies also need to learn how to navigate this ever-changing landscape where much remains unknown, Eeman said.

"We're making big bets in an environment that's changing all the time, and where many of the critical questions remain unanswered."

***

WAN-IFRA Members can watch Ezra Eeman's presentation on our Knowledge Hub.

Teemu Henriksson

Research Editor

[email protected]