Signal & Trust – Issue X

🔍 Go Deeper: Medill’s State of Local News Report 2026 includes county-level maps and data and Francesco Marconi’s Who Will Monetize Truth? examines the business case for selling intelligence and the funding challenge for public-service journalism.

signal scan

A curated round-up of how AI is reshaping news – and trust

Meta to Delay New Commercial Terms Criticised by Publishers, Advertisers

Just days before it was to take effect, Meta has delayed plans to scrape and reuse content from any site tied to a Facebook or Instagram business account. The proposed move has alarmed news publishers, regulators and advertisers.

Source: Reuters

In the AI Age, Publishers Who Understand What Readers Want Will Do Best, New Report Argues

A new Tow-Knight Center report argues even scoops can lose economic value quickly when AI repackages them. Publishers now face harder choices about what to cover – and what makes their work indispensable.

Source: Nieman Lab

Guardian and Sun Agree on Ban for Deceptive AI Bots

Eight of the UK’s largest news publishers, joined by Conde Nast and Reddit, want a law forcing AI crawlers to disclose who runs them and what they’re scraping.

Source: Press Gazette

Inside the Rise of AI-native Newsrooms Run by Autonomous Agents

What does it take and what does it mean to run a newsroom operated largely by AI agents? The Reuters Institute asked founders of five AI-forward news outlets where automation works best so far, and where humans are still needed.

Source: Reuters Institute

Life after SEO: How Newsrooms are Adapting to the AI Search Shift

With Google search referrals to publishers plunging 40% in a year, according to Chartbeat, newsroom search leads explain why they’re now betting on brand and loyal readers over clicks.

Source: Reuters Institute

“A generation ago the news industry largely hesitated when the internet arrived and began to disrupt the news business. That was a predictable but profoundly misguided reaction. If we make that same mistake now, the worst use cases of AI will fill that void.”

-Tom Rosenstiel,
Author of The Next Journalism: How the Press Must Change to Serve Democracy

As AI draws search traffic from news sites, and hands budget-conscious executives a new rationale for cost cuts, Tom Rosenstiel argues that journalism must quickly transition from an era of attention seeking to becoming more useful in people’s daily lives.  The author of the newly published The Next Journalism speaks with Aspen Digital about the difference between AI that improves reporting and AI that simply makes it cheaper and where newsroom leaders should start. This discussion has been edited for clarity and length.

Aspen Digital: In your new book you make the case AI is the newest technology publishers are hoping will help them reverse decline. What’s the risk that newsrooms treat it as another quick fix instead of confronting deeper problems?

Tom Rosenstiel: We used to practice a journalism of attention because in the old economic model of news we got the audience’s attention and then rented that attention to advertisers. Newspapers derived 80% of their revenue from ads, TV (prior to cable) 100%. And – the data is very clear – about how we got people’s attention. We did stories about things that were wrong (the watchdog barking) and things that were weird (man bites dog). Study after study shows those are the two dominant news values. They even dominate how we do coverage of politics and government. 

Meanwhile, as journalists we didn’t have to think all that much about how to make our stories useful or even what information would help people live their lives. If anything, we tended to look down at “service” journalism. And as newsrooms shrank, we did less of it.  That won’t work anymore. In a world in which most of the revenue must come from consumers, or members, or donors, the news itself must become indispensible. That means we need to think about what job each story is doing to help readers live their lives and improve their communities.   If anyone thinks AI is a quick fix to that conceptual shift, they will only accelerate journalism’s decline. 

Aspen Digital: You describe the long-promised shift from journalism as a product to journalism as a service as finally underway, built around giving people solutions and what they actually need. How does AI help publishers make that shift?

Tom Rosenstiel:  AI can be part of making that shift, but only as a tool, not as a singular solution. The shift needs to be a cultural and conceptual one in newsrooms, not merely a technological one. If publishers need a shorthand simple rule about what AI is going to help versus what AI is not, I would offer this: AI tools that are designed to help journalists make their reporting more comprehensive and accurate – that improve the quality of the news product for the consumer – are good. AI tools that are designed to remove people from the reporting and writing process primarily to make it cheaper, will only weaken an already inadequate product. 

What is an example of a tool that makes the reporting better? At the University of Maryland we have built tools that can monitor government meeting transcripts and live streams looking for stories or issues or decisions that journalists want to track. Suddenly, we can have 100 civic meetings monitored in the state every week that were going unmonitored and alert a reporter to anything the reporters or editors had deemed relevant beforehand. Other tools can scan interviews reporters have done against fact checkers and other digital resources to see if those interviewed made statements that were demonstrably false. 

What’s an example of a bad AI tool? Write my story for me. Do my interviews for me. Build a newsroom for this town that has no reporters or editors but pumps out machine produced stories with bylines of people who are not real and pretend you have built a news organization.

Aspen Digital: You’ve pointed out that when AI answers questions directly, a newsroom’s archive can become more valuable, not less. How should publishers be thinking about their back catalog now – as just a library that can be accessed more easily with AI or something more?

Tom Rosenstiel: There are a few things. One, which we are seeing already, is for newsrooms to build a bespoke AI archive of their work and then build chatbots in which members of the public can query the newsroom about the news and answer their own questions. It needs to be done well. The builders of those tools need to build them in such a way that they have tried scores of prompts, asked the chat bot every question they can think of, in effect road testing the bot. And cleaning up hallucinations. It’s not as simple as it might seem. It only takes one false answer for someone to dismiss that product as a disaster and to tell their friends and family not to trust it. Another example of how to use archives is internal. We in journalism test the accuracy of our work story by story. Did we spell the names right? Did the story seem fair? Done. But we do not go back very often over a year or two to assess how we have covered this topic or this community? Have we done so fairly, comprehensively? Have we represented the reality of that topic well? Have we covered that community holistically and depicted it accurately? At Maryland, again, we are building a tool that newsrooms can use to do just that.

Aspen Digital: Recently, McClatchy laid off more than 90 journalists while pushing an AI tool that repurposes reporters’ existing work into new stories. Your book is hopeful about AI, but on the ground it’s often the justification for cuts. How do you square the promise you write about with what’s clearly happening to jobs?

Tom Rosenstiel: Nowhere am I hopeful that AI will be a justification for job cuts. I advocate that we rush to build AI tools that can improve the journalism, not make it cheaper. The reason we need to move quickly to build AI tools to strengthen our newsgathering is so we can fend off those who are building tools, and trying to sell them to the news industry, purely to replace people. If we wait, or hesitate, that is exactly what will happen. 

A generation ago the news industry largely hesitated when the internet arrived and began to disrupt the news business. That was a predictable but profoundly misguided reaction. If we make that same mistake now, the worst use cases of AI will fill that void. We need to fill the void with AI that helps. Or it will be filled with AI that may create a civic holocaust in a lot of communities.

Aspen Digital: At Aspen Digital’s recent “Publishers, Platforms and the Future of News” convening, one worry kept surfacing: AI is becoming a personalized layer between audiences and the newsrooms that pay for reporting, answering the question so readers never reach the source. If that becomes most people’s front door to news, what happens to the outlets doing the original work?

Tom Rosenstiel: The only way I think that AI becomes a main entrance to a news organization would, again, be a chatbot of a newsroom’s archive. And it better be as close to accurate and bulletproof as you can make it. Do it badly and you hurt yourself. But that is not the only entrance to the news product. A successful news organization has to be a house with many doors. It might be high school sports. Or databases about education. Or topic pages that can catch me up on the news (imagine a wWikipedia page that is updated whenever a new event happens). Or it could be Q&As on key topics. It could be recipes. For many daily New York Times readers the front door is the games. 

Yes, AI is harming traffic. The solution is to build news products that go beyond the classic news story about what happened yesterday – the basic alert or update story – and start producing content that helps people live their lives. Some of that content will be conventional news stories. More of it than in the past will be content that is not so ephemeral, that is more evergreen, and that is rich enough that a crappy Chat GPT or Gemini summary is no substitute for the thing itself.  Journalism no longer needs to be simply what happened yesterday. That is an artifact of another technology.

Aspen Digital: For the news leader who’s convinced and wants to act on AI, what’s the first thing they should do?

Tom Rosenstiel: The book is about how journalism needs to change conceptually and walks through 10 different important ways to do that. There is one chapter in which I talk about AI, and that mostly addresses how newsrooms should take public data from their communities and try to build what I call a “civic internet” of data that people can navigate on their own.

Given that, what is the first thing a newsroom could do if they are curious about AI and have done nothing? Identify things that are time consuming for reporters that AI could help them with. That meeting tool I talked about is one. So are those tools that help reporters analyze interviews they have done. Some of these tools can now be built in a day. If there is a journalism school in your region, there are almost certainly faculty building some of these tools already. Ask them what they have. If you know of a newsroom that is using a chatbot that analyzes their archives, find out more about it.  So what is the first thing they should do? Take one step in the direction of finding a tool that will help you do your job, not replace you doing your job.

in the feed

One share worth a closer 👀

Media analyst Doug Shapiro argues that AI will weaken the link between time spent with content and the value people get from it because anyone can now compress a long piece to suit their purpose. To prove it, he had Google’s AI tools turn his own essay into a 13-page visual summary.