What happens when AI starts building the newspaper page?
AI in publishing is often discussed in terms of creating content. But publishers are also looking at where it can reduce the work involved in producing the newspaper.
That matters as print circulation continues to fall and publishers look for ways to make print production more efficient.
Print circulation and advertising still account for 43.6 percent of publishers’ total revenues, according to our World Press Trends Outlook 2025-2026. At the same time, 93 percent of respondents identify AI automation as a top investment for the years ahead.
During our recent WIZONE webinar, representatives from Atex, StoryEditor and EidosMedia showed different ways AI can be used in print production: suggesting headlines, placing stories, fitting content into templates and building complete pages.
The first step is deciding what should be automated
“AI suggests, editors decide,” said James Hirsz, Product Lead at Atex.
An editor prioritises a story, specifies its position and width, and lets the automation engine work out the rest of the page. AI then enters at a more specific point: when a headline is too long for the available space.
The system can suggest alternatives using the publication’s editorial style book. But it’s the editor’s decision whether to accept one, write a headline manually or ignore the suggestions, Hirsz said.
Among print’s constraints versus digital is finite space.
“A headline has to fit a box without damaging the page. A story has to fit the available area. Images have to work within the layout. Automation therefore has to understand both content and the rules governing its presentation,” Hirsz said.
For Atex, that means grounding AI in a publisher’s own style guide rather than relying on a generic model.
If automation is going to work in production, publishers need to make their own rules usable by the system.
“AI does not make the decision the editor does. The distinction is what makes the efficiency gain trustworthy,” he said.
Hirsz’s three recommendations point in that direction:
- Start narrow. Prove it, then expand. Don’t try to automate everything at once.
- Fix your content model before you go live. Automation is only as good as the content feeding it.
- Give AI the rules of your publication and ground it in your own style guide, not a generic model.
Automation gets harder when the page becomes the product
Publishers want print production to require fewer people, while the pages themselves can be more complex than a simple headline, image and block of text, according to Marko Margeta, Director of Product Development at StoryEditor.
StoryEditor approaches this through templates and predefined snippets rather than asking AI to invent page designs.
“If you give everything to AI, I think the result could be sterile or plain,” Margeta said.
The alternative is to automate within a system that already contains the publisher’s design logic. Articles can be connected to snippets and positions, with the system selecting templates according to factors such as article size, section, images and information boxes.
AI can also handle production constraints such as articles that are too long for the available space.
Instead of rewriting them, the system removes material it considers unnecessary so the story fits. The editor can review what was deleted, he said.
“The goal is not to remove every decision from the workflow. It is to reduce the number of decisions that require a person to intervene manually,” he said.
From assistance to autonomous page building
Jörg Drees, Director Marketing and Business Development, EidosMedia, showed how AI can generate complete print-ready pages from stories already prioritised by the newsroom.
Editors decide what runs, what leads, what is held and what matters to readers. The system handles placement, sizing, headlines, images and fitting the content to the page.
“That’s production work, not editorial work,” Drees said.
The system also learns the publication’s design. Rather than imposing a generic template library, EidosMedia analyses a title’s own production history and page archive, including the layout decisions made by its editors.
That addresses one of the central problems with automating print: a newspaper’s identity is not just its content. It is also how that content is arranged, he said.
“That matters because every newspaper has its own visual identity, its own way of handling a lead story, its own headline hierarchy, its own relationship between text and image on the page,” Drees said.
EidosMedia’s system is running across more than 60 titles, with more than 46,000 pages produced in 2026 and more than 12,000 hours of layout and quality-assurance time saved.
Those numbers matter because the test for print automation is not whether it can produce an impressive demonstration. It is whether it can survive the deadline pressure of a real newspaper: late images, changing stories and the need to make final decisions quickly.
What publishers need to change before automation works
A publisher cannot simply add AI to a workflow that was designed around manual production and expect the process to become efficient. The underlying content, templates, priorities and rules have to be structured well enough for automation to use them.
Approaches may differ in implementation but not in their basic logic.
In each case, automation depends on decisions that publishers have already made about how their products should work.
This also changes the meaning of “AI automation.”
It does not necessarily mean handing the whole production process to a model. It can mean turning a publisher’s existing rules, templates and production history into something software can execute.
The payoff is not just fewer layout hours
The strongest case for automation may be what publishers do with the time saved.
Drees described two uses for the time recovered: quality and flexibility.
If editors are no longer spending their days fitting stories and images into pages, they have more time to look at the finished product and ask whether the right image was used, whether the headline hierarchy works and whether the front page deserves more attention, Drees said.
The other benefit is a more flexible production process.
“If layout takes minutes rather than hours, publishers can leave decisions later, accommodate breaking news and produce additional sections or special editions without adding the same amount of manual work,” he added.
However, all three presentations made the same point: editorial judgment remains with people.
As Hirsz put it, “AI does not make the decision, the editor does. The distinction is what makes the efficiency gain trustworthy.
Too short? WAN-IFRA Members can watch the full session on our Knowledge Hub.
