This article was originally published by Northwestern University’s Medill Local News Initiative and is republished here with permission.
In April 2024, the Local News Initiative, in conjunction with the Knight Lab at the Medill School of Journalism, Media, Integrated Communications and Konrad-Adenauer-Stiftung USA, published a report titled, “Impact of AI on Local News Models: AI Is Disrupting the Local News Industry. Will It Unlock Growth or Be an Existential Threat?”
Based on discussions with more than 25 local news and AI experts worldwide, the report explored the potential benefits and perils presented by this revolutionary technology. University of Maryland journalism professor Tom Rosenstiel, a participant in the Washington, D.C., workshop that yielded much of the report’s content, urged journalists to take a proactive approach to generative artificial intelligence and to avoid the industry’s mistakes from a generation earlier as it was crushed by the emergence of the internet.
We checked back in with Rosenstiel, the Eleanor Merrill Scholar on the Future of Journalism at Maryland, for a progress report about AI and the local news landscape. These topics are top of mind for Rosenstiel, who is finishing work on a book titled “The Next Journalism: How the Press Needs To Change To Serve Democracy in the 21st Century,” due out in July 2026.
This conversation, conducted over a video call, has been edited for length and clarity.
Mark Caro: In our AI report, you say, “When the internet happened, we were basically in fear and denial. Now we have another chance, because this is as big as the internet.” How are we doing with that chance?
Tom Rosenstiel: That’s a good question. It’s a little hard to know. We see glimpses that are really encouraging, but we also hear about incidents that are discouraging. If you go to CalMatters, they are using AI to track all of the committee hearings and all the transcripts of all the committee hearings in the California state legislature. Not only are they using AI in that very powerful way to monitor things that they could never have enough people to do manually, but they’ve created a website where I, as a user, can go and search any topic I’m interested in, and AI will find the conversation that was had in the state legislature about that topic and pull those transcripts for me. So it’s an impressive tool.
To me, that’s a sign that people can do it and illustrates, “What does ethical AI use in journalism look like?” One fear I have is that not that many people probably know about it. I don’t know how many people have used it, and I don’t know to what degree it has been replicated by other news organizations in other states. I certainly haven’t seen it rolled out elsewhere, and that makes me worry that we’re kind of back in the world of everybody’s doing their own thing, inventing their own jalopy in their garage.
On the optimistic side of the ledger, there’s a lot of AI money to be had. If you say you want to do AI, the foundations that care about journalism seem quite interested in that. It’s the flavor of the month, rather than, “Oh, well, journalism is under siege. I’m not going to put my money there. I’m worried that that money is going to be squandered.” On the worrisome side, I have the sense that everybody’s doing things to the extent that they have resources but that there may not be as much banding together or gathering as might be required.
Caro: In our workshop, people thought it would be great to use AI as a tool to gather information, to process data and to put stuff together that is really hard to put together manually, but we don’t want it to replace us in doing the interpretation and writing the stories. Then you get what happened at the Chicago Sun-Times and Philadelphia Inquirer, which published an outsourced “Heat Index” section featuring an AI-generated summer reading list in which most of the books didn’t actually exist, were AI hallucinations and weren’t caught by the piece’s author.
Rosenstiel: The line that I’ve heard — it’s a little glib, but it’s memorable — is that AI right now is good for aggregation but not creation. It’s great at trying to find all the things in this massive data set that will answer this question, and then you can check that. It’s the same thing that we’ve known about computers for decades: They’re really good at counting; they’re not as good at creating. What are the things that journalists aspire to and that people dislike us for because we’re not good enough at? Being fair, getting nuance, creating the right context, putting things in a larger perspective. That’s not counting.
Caro: Part of our fear is we don’t want robots to replace us, and we don’t want businesses that don’t care as much about nuance to think they could use fewer reporters with AI. When something like this fake reading list thing becomes national news, on one hand, it’s terrible and embarrassing. But on the other hand, maybe that’s a helpful lesson: Guess what? We still need the humans to put this stuff together, because this is what happens when you rely on the robots.
Rosenstiel: I think that when you have these incidents, some people are going to go, “Oh, great. I don’t have to worry about AI yet.” Go back to 2000: “Oh, the internet bubble happened. Phew, we missed that. Don’t have to worry about that.” I think it’s critical to learn and advance from these mistakes and not use them as an excuse to say, “Good, I can put that off.”
Again, this is an oversimplification, but if you’re using AI to try and figure out how to make your journalism better, that’s good. If you are using AI to try and figure out how to make your journalism cheaper, that’s bad. Because cheaper is going to be worse, and it’s going to hurt all of journalism, because it’s going to make people think, “What’s the point of this? This product is no good. It’s not helping me. It’s a joke. It’s not what it used to be.”
Caro: At the workshop, you said, “Journalism must shift from being an alarmist watchdog to ‘How do I help you make your life better?’ Have you seen evidence that this is happening?
Rosenstiel: Yes, some. I think that CalMatters’ use of AI is a good example of that. Much of what made journalism in the 20th century indispensable for people was the stuff that surrounded the journalism. You didn’t know what was at the movies or what was on television. You couldn’t sell your bike or buy a house or find an apartment. You needed the package that came with the newspaper to live your life. We the journalists didn’t think too much about how our journalism was helping people. Now we do because most of that indispensable stuff people can now get elsewhere. To make journalism sustainable, we need to make our journalism something that helps people live their lives.
The one thing that vanished from newspapers was calendars, right? Because all of a sudden, if I wanted to go to the theater, I’d just go to their website and see what was playing. If I wanted to go to a ball game, I would go to their website and see if I could get tickets. So we lost the kind of centralized community calendar. Actually, AI could reproduce that now, and it might be better than my having to go to eight different places on my own.
One of the early ideas around digital that didn’t really go anywhere was can we create a digital Main Street so that local retailers who are being creamed by Amazon can have a fighting chance? Can a local news community news organization create a virtual downtown where I could go to the news site and see what’s in the local stores, support those stores, buy those things from those stores? That was beyond the capability of local news organizations 20 years ago. AI may make the idea of creating a virtual downtown that will help local merchants plausible. That’s also a revenue source, because if I buy it through that virtual downtown, the newspaper, like Wirecutter, can take a small percentage of the sale.
So there are options, I think, for AI not just to help us make the journalism better by taking over tasks that are tedious and time-consuming for reporters. Are there new revenue opportunities, even in ideas that were considered in the first generation of the internet but were impractical?
Caro: There’s the AI that can help you make money, and then there’s the AI that everyone’s afraid is going to gobble up all the money. Multiple states are now seeking compensation from Google and Meta for their content. What are your hopes for local news organizations being able to get some revenue back from Big Tech?
Rosenstiel: Well, that’s a good question, and that’s a political question, and of course, if it’s political, the answer varies depending on where you are, right? If you’ve got a Democratic governor and a state that wants to contrast its policies with those of President Trump, they’re going to be friendlier to the news media. And if you’re in a conservative state where the governor is trying to mimic Trump’s antipathy towards the press, that seems like a much more uphill fight.
Another threat along those lines with AI is that I go to AI search, and I ask a question. It gives me the answer in a synthesized way, so it drives no traffic to any of the news sites. Now tech is not just ripping off the headline and the links, and I can peruse those and get my answer. They’re actually ripping off the intellectual capital of what I said, and AI is summarizing it.
Caro: That was discussed at the workshop as an existential threat, the idea that AI is going to eat up all the traffic.
Rosenstiel: It could, I mean, it’s an existential threat to Google, which is why Google’s going to try to sell ads against its AI summary, because it’s not driving people to links. It may be that people don’t even go to Google, because there’s some other AI thing that just answers the questions for me, and I don’t need search links. So Google’s under threat from that, too.
What are some potential answers or responses for the news business? We should be summarizing our own material. In other words, you want to know the answer to something that happened in Chicago? What if the Chicago Tribune or the Sun-Times had a really superb AI answer machine to create what’s essentially a bespoke Wikipedia answer to my question? Or I go to The Washington Post and say, “I’d like to know the history of this neighborhood. I’m thinking of buying a house there. Give me the crime statistics for this neighborhood. Give me the ratings of the schools for this neighborhood.” So that the archives of a local news organization — or a group of local news organizations that bind together — becomes a source of local intelligence. That is really useful and perhaps worth paying for. Maybe not a lot, but maybe something.
Caro: Do you think there’s a way these news organizations could create those answers and protect that information, as opposed to AI companies scraping that information?
Rosenstiel: Yeah, I think there are ways. We’re speculating here, but this could be exactly the opportunity of a second chance, right? What happened the first time around was Google was scraping the leads and headlines of all the news stories, and their claim was, “We’re driving traffic to you. We don’t have to pay you.” There was some discussion among news organizations about, “Why don’t you just remove your stuff from Google? Why don’t you just say to Google, ‘You can’t have it’?” Well, it was too risky a bargain to make, right? Google wouldn’t really negotiate with small guys, and the big guys were like, “Well, we do get a fair amount of traffic from them, and I don’t think we want to cut them off entirely.” If Google is in the same boat we are now, where AI may make search irrelevant, why not skip over that and create essentially a news answer machine that is the news organizations themselves binding together, putting their archives there and saying, “You can be paid for this”?
Another iteration of this is something called versioning. Versioning is “I am curious about tariffs, but I’m not that interested and haven’t read any other stories about them, but I kind of want to know what’s going on.” Because the computer knows you haven’t looked at any tariff stories from us before, we’re going to give you a very basic version of it, and AI can do that. AI can essentially draw from our latest story but give you simple background versus “Oh yeah, I see you’ve read all our tariff stories. Here is the latest, and it’s pretty detailed.”
The news has always seemed like a party where I don’t know anybody, I show up late, I don’t know what the conversations are about, it’s a lousy party, and I leave because I’m a little lost. AI has the potential to help solve that problem if we lean into that enough. That’s not a problem that a four-person nonprofit is going to solve, but the problem that that four-person nonprofit has regarding this is the same as The New York Times has, and if we could solve those together, it could be powerful. I really do think the threat to Google is an opportunity for the news industry.
Caro: The idea that AI is going to know what you’ve read and will come up with versions for you, on one hand, is helpful. On the other hand, knowing that there’s this entity that’s keeping track of what you’ve read — or at least what you’ve looked at and how long you spent on each of these articles — accelerates the whole Big Brother notion.
Rosenstiel: Yeah, we’re already there. Google literally knows every web page you’ve ever been to. You know they know your entire digital history. I don’t know what they’re doing with that information, and whatever they say they’re doing, I’m not sure I trust them — and I trust them more than I trust a lot of other companies. If it were a news organization doing it, I think you could have more trust. We’re not selling your data to anybody. We’re only using it to make this product more useful to you because our entire business model exists around being useful and being trusted. So if we do anything else with it, we’re cutting our own throats.
Caro: Right now, the press is under attack more than ever and less trusted than ever, in part because of these relentless attacks. Outlets are closing. News deserts are expanding. The National Trust for Local News has been selling off its papers. The governor of California has gone back on its deal with Google and reduced the state’s first-year contribution to local news organizations from $30 million to $10 million, prompting Google to cut its contribution from $15 million to $10 million. Where are we now with the state of local news? Is this a crisis with a capital C?
Rosenstiel: First, local news is the product that failed in the marketplace, right? It was often a weak product. It really depended on these journalism-adjacent elements to be indispensable. The journalism was often inconsistent, not very good, and so local news is in crisis. The New York Times is succeeding because it’s innovated its product, and it’s no longer failing in the marketplace. So local news is the product that needs to innovate most.
At the same time, local news is the arena in which there’s the greatest potential to create a common public square and common facts. Why? Because the data is very clear: It is more trusted than national news. There is more commonality in the audience, particularly of local television news, across party and ideology. You’ve got more people who disagree with each other watching the same local newscasts or reading the same local newspaper than you have in the ideologically fragmented national media ecosystem, and we have more common ground. We root for the same teams. We suffer the same weather, etc., etc. So that’s the world in which we still have shared common facts.
That’s also where local retailers are suffering, so businesses have a similar incentive to try and ally with a local publisher. That’s why I think journalism has to sort of reorient itself from, “We’re monitoring the powerful on your behalf and doing once a year maybe something semi-investigative” to, “How can we understand you better and help you live your lives better at the local level?” All of our journalism really needs to be service journalism. I don’t mean that we stop covering city hall. I mean that we cover city hall in a way that I understand. What are the things that you need to know about city hall? What’s relevant to you, and how can I cover this in a way where I don’t just tell you these two guys are fighting, you’ve never heard of them before, and this guy got the better of that other guy? No, I’m going to tell you they’re coming up with this plan, and here’s what the plan would do, and here’s how it would affect you, and if you want to get involved, here’s how you can do that.
There is a way to make our accountability journalism into service journalism, but we really need to get closer to our audience to do that. It’s not an easy task. But the crisis is its most acute at the local level, and the potential is greatest at the local level. I think that the effort to put money into and try and focus on local news has been well-intended but not well-executed. I think it needed to start much more with: What were the flaws in the product, and how can we rethink and reinvent them?
Caro: Expanding news deserts and ghost papers, where chains have eliminated local reporting, push people to get their news from national sources. So local news has to get people reinterested in what’s going on down the street, right?
Rosenstiel: The local paper had the advantage that it was the portal to everything. The Chicago Tribune had all the wire stories from everywhere anyway, so my local news product was my window into everything. It was my department store for news and for everything else. Once it became very easy to get national news elsewhere, that was yet another thing that I didn’t need my local news outlet for, just like I didn’t get classifieds from it, and I didn’t get TV listings from it.
So most local publications said, “Well, OK, let’s be hyperlocal, cover only things that local news can cover.” But they were defining local news in a very traditional way: Let’s go cover this building. Let’s go cover this. Let’s go cover that. They weren’t thinking conceptually about how do people live their lives? We were still defining beats, like the police beat and things like that, and when you go into a newsroom, as I have, and say, “Well, let’s have a sense-of-place beat. What is it like to live here?” journalists are like, “What? How do you do that?”
Or, “How is this place changing?” Those are the ways that people talk about things. Or, “How could the city be better?” You could have a beat that was basically how to make the city better. It might work. It might not work. You need to dive in and do research in a given community. But newspapers or news organizations have the capability of doing that. They have ethnographic researchers who could go into the community and find out what it is people need to know. Those people are called reporters, and they know how to interview people, but we don’t tend to do that. We actually have the skills to try and get closer to our community.
Comments