The discussion has shifted to a more fundamental question.
What role will journalism play when AI increasingly mediates how information is discovered, consumed and distributed?
That shift was visible across presentations from publishers, news agencies and technology executives throughout the three-day congress.
Two years ago, much of the industry's attention focused on newsroom automation, content generation and workflow efficiencies. Those topics remain important. But in Marseille, the most consequential discussions centered on user experience, personalization, knowledge architecture, audience loyalty and the relationship between journalism and AI agents.
The focus was no longer on how AI could help create content.
It was on how audiences would experience journalism in an AI environment.
The Hindu Group in India provided one example of this transition. The company has been experimenting with presenting the same journalism in multiple formats, including AI-generated summaries, question-and-answer formats, audio explainers and alternative article lengths. The objective is not simply greater efficiency but greater accessibility and discoverability.
Sweden's Bonnier News is pursuing a similar goal through conversational archives. Rather than relying on traditional search, readers can query decades of reporting through natural-language interfaces. The shift is from retrieval to conversation.
At Scroll.in, the focus is on contextual understanding. The publisher is developing AI-powered workspaces that combine timelines, event clusters, knowledge graphs and dynamic FAQs, particularly for researchers and specialist audiences seeking comprehensive understanding of complex topics.
Different organizations are pursuing different approaches, but they share a common assumption: the article is no longer the only product.
The most forward-looking example presented in Marseille may have come from Germany's dpa.
Astrid Maier, chief deputy editor and head of strategy, introduced dpa IQ, a system designed to make journalistic information directly accessible to AI agents and workflows.
"The news artifact is not the end. It's only the beginning," she said.
The premise behind dpa IQ is that journalism can be structured as machine-readable knowledge rather than distributed solely as articles. AI systems can retrieve facts, timelines, contextual information and archived reporting through APIs and agent frameworks.
The significance of the project extends beyond technology.
It represents a different understanding of what a news agency might become in an AI ecosystem: not merely a distributor of articles, but a provider of trusted information infrastructure.
Austria's Kleine Zeitung offered another perspective on the changing media environment.
Sebastian Krause, the publisher's head of digital, argued that publishers must begin thinking about AI systems as a new category of audience.
For years, publishers optimized content for search engines. Increasingly, however, AI assistants are reading, summarizing and retrieving information on behalf of users. This shift has prompted discussions about Answer Engine Optimization (AEO) and Generative Engine Optimization (GEO), reflecting the growing importance of visibility within AI-generated responses.
The implication is significant.
As AI systems become intermediaries between publishers and audiences, the strategic challenge is no longer simply generating traffic. It is maintaining visibility, attribution and commercial value in environments where users may never visit the original source.
Yet for all the discussion surrounding AI, one of the strongest messages of the congress concerned journalism itself.
The most widely discussed speech of the week came from New York Times publisher A.G. Sulzberger.
Sulzberger argued that while AI companies can organize, summarize and distribute information, they remain dependent on journalism's most fundamental contribution: original reporting.
Facts do not appear spontaneously in AI systems.
They originate from reporting in the field, interviews, documents, investigations and verification.
As AI-generated content becomes increasingly abundant, speakers throughout the congress repeatedly returned to the same conclusion: trust, verification and firsthand reporting become more valuable, not less.
This may prove to be one of the central paradoxes of the AI era.
The more powerful AI becomes, the more important journalism's uniquely human functions appear.
For South Korea, where AI adoption is among the fastest in the world, the discussions in Marseille offered an important reminder.
Automation, translation and workflow efficiencies are necessary developments. But they are not, by themselves, a strategy.
The organizations shaping the industry's future are increasingly focused on audience relationships, user experience, information architecture and trust.
The debate is moving beyond how AI can improve newsroom operations.
It is moving toward how journalism itself must evolve.
As WAN-IFRA AI lead Ezra Eeman observed during the congress, "There is no map."
That remains true.
But the direction of travel is becoming clearer.
The conversation is moving from content to experience, from search to interaction, from distribution to loyalty, and from information abundance to trusted verification.
In that sense, the most important discussions in Marseille were not ultimately about artificial intelligence.
They were about the future value of journalism.
Copyright ⓒ Aju Press All rights reserved.