Before ChatGPT was launched more than two years ago, AI existed mostly in the world of data scientists and specialists.

“It was complex, technical, and quite removed from our daily work in media. Then came this simple chat interface, and suddenly, everything shifted. AI stepped out of the lab and into everyone’s life, bringing with it a wave of new models and apps,” Ezra Eeman, AI Expert and Director of Strategy & Innovation at Netherlands-based NPO, told the audience at our recent AI Forum in Bengaluru, India. Eeman is also a strategic advisor for WAN-IFRA’s AI in Media initiative.

Media has evolved from traditional formats (newspapers, cable, and broadcast) to digital media, and now to AI-driven media, where hyper-personalised content can be generated in real-time.

While traditional media had fixed formats, digital media enabled segmented and responsive content. However, media augmented by AI is able to deliver fluid, adaptive formats that can engage individual audience members interactively.

Not just another digital tool

“We have to understand that AI isn’t just another digital tool in the box. It’s actually changing the box itself,” Eeman said.

AI creation tools are now integrated into newsroom systems. Today, AI is being used for creating headlines, summaries, news edits, transcripts, translations, video clipping, and more.

AI tools are becoming as essential as the tools we’ve used for decades, and newsrooms are relying on them for tasks that were once time-consuming, such as generating summaries and transcripts or producing breaking news in real-time.

AI is also enabling anyone to produce professional-looking content – tools that were once reserved for newsrooms are now accessible to anyone. The steps in content creation are also shrinking. Previously, specific skills and teams were needed for each step, but now many of those steps are automated, Eeman said.

For instance, Switzerland-based Tamedia has reduced the time required to produce newsletters with the help of AI. Meanwhile, Schibsted is using AI to speed up transcriptions. In the UK, Reach has cut the production time for breaking news from nine minutes to 90 seconds.

What exactly is ‘fluid’ media?

AI is moving media from traditional linear flows into a “fluid” state, according to Eeman. Content can now adapt in real-time to individual preferences, remembering past interactions to create personalised experiences.

• Dynamic content creation: Stories can evolve in real-time as new data is fed into them. For example, Dataminer, a company specialising in early warning signs, uses AI to update story descriptions as events unfold.
• Adaptive platforms: Facebook’s “Imagined for You” prototype creates a personalised news feed based on a user’s interests. Google’s adaptive interfaces change based on user needs, generating everything from weather maps to visual storybooks.
• Interactivity and AI conversations: AI enables dynamic conversations, such as the interactive news podcast by Hume AI, where listeners can engage with the podcast’s content in real-time. This adds new layers of interactivity that are still being explored.

(left) Facebook’s ‘Imagined for You’ prototype creates a personalised news feed based on user preferences. Google’s adaptive interfaces change based on user needs

AI agents: Next wave of autonomous systems

The next phase of AI in media involves AI agents (AI systems that can independently execute tasks).

Beyond responding to prompts, these agents can perform jobs such as building and posting videos or curating newsletters autonomously, Eeman said.

A potential application of AI agents is a personal newsletter that scans favourite news sources, curates content, and suggests it to the user.

However, these tools are far from perfect. “AI models are prone to errors, such as hallucinations, slow responses, and incomplete or inaccurate results,” Eeman said.

“There are significant concerns related to privacy, copyright, and bias. It’s crucial not to replace authentic voices with content that sacrifices quality in favour of speed,” he added.

AI enables scale and variation in distribution

Distribution is another key area facing transformation. Compared to traditional search engines, AI can extract key information from various sources and synthesise a complete answer directly in the interface, rather than listing different sources of information.

“This is happening in all AI interfaces. Companies like Microsoft and Apple are integrating it into their operating systems,” Eeman said.

Apps such as Particle News can also create smart summaries that strip away the original reporting.

There are AI models building complete product interfaces on top of models such as Perplexity, which creates full long reads that closely resemble premium original stories from publishers.

Social media apps are also integrating AI chatbots into their services, serving up news, entertainment, and sports – traditional territory of publishers.

AI in distribution enables scale and variation. One well-researched article can be transformed into hundreds of versions, each tailored to different audiences and contexts. The challenge now is not scaling, but maintaining quality, relevance, and trust at that scale, Eeman said.

Publishers need to differentiate themselves

With the rise of digital and AI, the value chain in media has shifted. What was once controlled by publishers is now in the hands of external players – cloud infrastructure, app stores, and social media platforms.

Generative AI has also placed the newsroom process under pressure. In its most radical form, AI can bypass traditional steps, going directly from data to interface, shrinking the newsroom process.

This shift is also changing where the power in the value chain lies. It’s about data, intellectual property (IP), talent, and the interface – because that’s where value can be monetised.

Currently, publishers hold strong positions at both ends of the value chain, with vast archives of quality content, original voices, powerful brands, and destinations.

But AI companies have been training their models on publishers’ content online for years. Bot activity continues to rise, with scraping content up 32%, according to Eeman.

Publishers have limited options: block access, take legal action, or partner with AI companies, increasingly seen in licensing deals.

These deals fall into three categories:

• Training licenses for access to archives (one-time or annual fee)
• Inference licenses (per-use fees), and
• Ad revenue models, which are more complex and not fully mature.

These deals are not open to all media organisations. Larger publishers have the scale to negotiate direct partnerships and custom integrations, while smaller publishers often face take-it-or-leave-it terms.

For smaller publishers, collective action may be necessary. Understanding where you stand in the new landscape is key, and publishers need to differentiate themselves. In an AI world, it’s not about being better or faster – it’s about standing out, Eeman concluded.