NEW YORK — The era of the static article is effectively over. As newsrooms grapple with fragmented audiences and the rise of agentic AI, the fundamental unit of journalism is shifting from a fixed text artifact into what industry futurists call “liquid content” — information that instantly adapts to the user’s context, format preference and platform.
At TVNewsCheck‘s NewsTechForum 2025 panel “How Will AI Revolutionize the News Product?” on Tuesday, technology leaders from CNN, The New York Times, Reuters and Hacks/Hackers detailed a radical restructuring of the news product. The consensus? The future isn’t just about automated writing; it is about building an information architecture where journalists become developers, archives become semantic databases, and the “website” may eventually become obsolete.
The Efficiency Engine: Unlocking The Archive
Before AI reshapes the front-end user experience, it is aggressively overhauling the back-end plumbing. For legacy organizations sitting on decades of content, the immediate value proposition lies in semantic search — the ability to query archives by meaning rather than keyword.
“From a broadcast and a video point of view, CNN is working on trying to unlock semantic search of some of our video archives,” said Ryan Struyk, director of AI innovation at CNN. He noted the shift away from relying solely on human-entered metadata to “unlock a more sophisticated semantic search” that allows producers to find the proverbial needle in the haystack instantly.
This utility extends to investigative journalism, where the volume of data often outpaces human capacity to process it.
“If you were to drop, say, 10,000 documents on the floor, AI would help to surface the 100 or so that you actually need to pay attention to,” said Rubina Madan Fillion, associate editorial director of AI initiatives at The New York Times.
She highlighted the Times’ internal tool, “Sheetsheet,” an AI-powered smart spreadsheet developed by colleague Dylan Freedman, which allows reporters to sift through massive datasets for investigations.
Quantifiable Velocity: The Production Shift
For Reuters, a 175-year-old agency built on the speed of information, AI has become a critical lever for operational velocity. The agency has successfully compressed labor-intensive production tasks, freeing up producers for higher-value editorial work.
“It took about three to five minutes for a talented producer to actually repackage a story for publication on our website,” said Arlyn Gajilan, global editor of AI development and integration at Reuters. Through AI-driven asset association, “we’ve taken that three to five minutes down to about 35 to 40 seconds.”
Gajilan noted that when multiplied by 400 stories a day, the efficiency gains are exponential. Furthermore, the agency is deploying an AI suite for broadcast clients that can auto-transcribe in 57 languages and translate into seven, alongside scene detection for B-roll — effectively automating the rough cut.
The Rise Of ‘Liquid Content’
The discussion pivoted toward a more existential shift: the concept of “liquid content,” where a single reporting asset is no longer tethered to one format. As audiences increasingly rely on AI interfaces for information, the news product must become fluid.
“We are entering into this [era] of … liquid content, where a single unit of artifact is no longer the exact place in which people are able to consume it,” noted moderator Nikita Roy, founder of Newsroom Robots Lab, during the session.
Fillion validated this approach, describing how the Times uses AI to monitor the fragmented ecosystem of the “manosphere” — a network of influencers and podcasts that traditional monitoring misses. The Times built a “Manosphere Report,” a daily AI-generated summary of podcasts and YouTube channels, to help reporters track narratives before they hit the national stage.
“It’s like, think about the context of where you’re about to get on the elevator at work. What do you need to know going up?” Fillion asked. “What format do you need that news delivered to you?”
The ‘Vibe Coding’ Culture And The Death Of The CMS
Perhaps the most provocative insight came from Burt Herman, co-founder of Hacks/Hackers, who suggested that the rigid Content Management System (CMS) might be nearing its end. He argued that as coding tools become accessible, the line between journalist and developer will dissolve.
“In five years, everyone in a newsroom will be a coder and using AI to build things,” Herman predicted, using the term “vibe coding” to describe rapid, AI-assisted prototyping by non-technical staff. “I don’t think people will have CMSes actually in the future. You’ll just build stuff that will create whatever you need it to be.”
Herman urged newsrooms to prepare for a “post-website” world where AI agents — not browsers — are the primary gatekeepers.
“I don’t think people are going to go to websites,” Herman said. “The AI agent … will go and sift through all that and just give it to you.” He advocated for the adoption of the “Model Context Protocol,” a standard that allows newsrooms to feed structured, local data directly into Large Language Models (LLMs), ensuring that when a user asks about a local school board, the AI knows exactly where to look.
The Trust Paradox
Despite the enthusiasm for innovation, the panel remained clear-eyed about the risks, particularly regarding breaking news. As distribution speeds up, the margin for error narrows.
“When it comes to a breaking news situation, that’s when the information that you might get from AI becomes just a little bit even more unreliable,” Struyk warned. He emphasized that while CNN innovates, “our journalistic integrity is everything to us. So complete accuracy is really the bottom line.”
The consensus was that while AI can handle archival retrieval and formatting, the “human in the loop” remains non-negotiable for factual verification. However, Gajilan raised a critical governance challenge.
“I often have to explain why we need governance on these things because in six months from now, this reliance on human in the loop … is not going to be viable,” Gajilan said. “How do we build an ethics framework that can match the speed with which we are developing?”
The Outlook
The trajectory is clear: Newsrooms are moving from static production to dynamic, AI-assisted generation. The value proposition for local and national outlets alike will shift from simply publishing text to managing rich metadata and building products that integrate seamlessly with the AI agents of the future.
“The future of media isn’t just about adopting AI — it’s about wielding it wisely and strategically,” Gajilan said. “If we don’t get [the metadata] right, you don’t get it right.”
About Jon Accarrino