Image

Rodrigo Zamith, associate professor of journalism, has published a new book exploring how artificial intelligence is reshaping journalism. 

“Journalism in the Age of AI: from Acceleration to Reimagination,” (Polity Press, 2026), co-authored by Seth C. Lewis and Tomás Dodds, examines everything from how AI is changing newsrooms to what the technology means for democracy. A digital download of the book is available for free at journalismandai.com.

Zamith recently joined the Office of News and Media Relations for an interview to discuss the book. 

The transcript that follows has been edited for clarity and length. An audio recording of the interview also is available below and on Soundcloud.

What inspired you to write this book?

I’ve been studying the intersection between journalism and technology for my entire academic career. I started with audience analytics—so how news organizations were trying to better understand what information people are interested in. Then, I studied data-driven news work—so data journalism and how news organizations were using computation to better serve some audience needs and maybe personalize some stories.

That has naturally led to understanding how this new technology that we have with us in generative AI—how it is both challenging the economic foundations of quality journalism, but also may be opening up some opportunities for news organizations and for journalists to do better journalism, and journalism that serves audience information needs better and in ways that they simply could not do so before.

In the book you compare journalism to a hamster wheel, explain that and AI’s role in that.

The hamster wheel metaphor is something that a journalist, Dean Starkman, coined back in 2010 or so to describe what he saw as accelerationism in the news industry. Basically, he wrote that the hamster wheel isn’t speed, it’s motion for motion’s sake. It is volume without thought.

Image

The idea there is basically that journalists were being asked to produce more and more content, and do so faster and faster, without necessarily being able to engage in the effort necessary to produce more quality. In one example we give in the book, The Wall Street Journal asked its journalists to increase their output by almost 70% over the space of eight years. That is unsustainable.

Many journalists expressed burnout with the level of volume that was being expected of them. And now AI comes along and it offers the possibility of accelerating that hamster wheel even further—by offering this new tool that allows news organizations to generate even more content than they ever could before with purely human workers. But is that necessarily translating into better journalism?

That’s the question at hand today. How can we take this technology that can be used to increase efficiency, and produce more but not necessarily better work, and try to re-orient the direction of its development and implementation in a way that serves the public better? It’s also a way of hopefully stepping off the hamster wheel so we’re not just producing more content for the sake of it, but maybe better understanding what is actually valuable to people and how we can provide more of that information.

You hinted at this earlier, but the economics of journalism are changing because of AI. Twenty or 25 years ago, things changed because classifieds moved online, and a major revenue source, for newspapers in particular, dried up. And now with these AI summaries you’re getting when you search for something, it doesn’t even drive people to the news outlets website anymore. So, does that take away some of the incentive of just quantity over quality?

I think it takes away some of the incentive because for many news organizations, I think the question becomes: “What is our unique value proposition when these AI summaries can already generate the basic information [about a story] faster and cheaper than what we can offer?”

That maybe does incentivize a repositioning away from simply providing a lot of superficial content and more toward providing deeper content that people would come to you for because they know that is what you have to offer, and that others are not offering. 

You pointed to the fact that we have been witnessing this economic disruption for over 20 years now. I think the key difference is that the platform era really focused on [tech] companies coming in and disrupting the distribution of news. People might go to social media or news aggregators for news, but what was happening there is that original news content was being shortened but still offered links to the original piece, if you wanted to learn more about it.

I think what is different in the age of AI is that these algorithms are now directly impacting the production of information. You now have [AI] syntheses that are being generated that address a specific information need that a news consumer might have. And what’s different now is that they might feel like their question was already answered by (a Google AI Overview).

As a result, (the news consumer) does not feel the need to click through to the original content. If they are asking ChatGPT or Claude a question about the news, they are getting a direct answer (to their question) that might include some  indicators about the source, but we know from the research that very few people actually click through to that original content. 

The implication there is twofold.

One, it means that fewer people are going to the original news content and, therefore, fewer people can be exposed to the advertisements that allow news organizations to monetize their work.

Two, it is increasingly severing the relationship between news organizations and their audiences because there is less interaction between individual audience members and the original (content producers) because these AI syntheses are already extracting a lot of the value from the news organization.

What that means economically is that it makes it harder for a news organization to develop the sorts of relationships with audiences that can turn into subscriptions and other ways of sustainably monetizing the work that those audiences appear to value.

As someone who teaches the next generation of journalists, what do you tell your students about AI use? How will future journalists differ from today’s journalists?

I think one of the main things we have to communicate to our students is that AI can make them feel like they are knowledgeable [about a topic] when they ask a chatbot a question and receive a response that makes sense to them. They can look at that response and just think to themselves: “Okay, this makes sense to me. Therefore, I understand the material.” 

[We must] help them realize that just because an answer makes sense to you, it doesn’t mean that you fully understand the material in the sense that you could come up with that answer on your own.

More broadly, it means that for some of the more superficial questions, AI might be able to produce a helpful answer to them. But when they start getting into to the more advanced tasks, the more challenging material that we cover in our classes, unless they know how to write a good prompt—which requires understanding the material on more than a superficial level—then they’re not going to be able to get the same quality of responses.

When it comes to future journalists, if you are just using very generic prompts and relying on (the same tools) to do the work for you, then everyone else can do the same thing [you can]. 

So, what is unique about what you have to offer? Why should anyone hire you instead of someone else who can also do the same superficial-level work, and is perhaps willing to do so for less money than you?

What I try to teach our students is that if you are serious about going into a highly competitive industry like journalism, you have to be able to offer unique value. And that goes beyond just knowing how to load up a chatbot and enter some simple prompts. 

It means understanding how to do a lot of the work yourself—so you can know when it is appropriate to use an AI tool, when it’s actually going to be helpful for you, save you time, and produce higher-quality work that will elevate you in the eyes of your employers. 

In your view, is there a red line, if you will, of AI use and journalism—things you should absolutely not do? I imagine perhaps creating a fake byline for a chatbot, something like that?  And has that line evolved as the technology has evolved and our acceptance of it has changed?

Absolutely. I think that when AI is being used in ways that meaningfully impact a news product, then that information should be disclosed to people. And there are some organizations that, in the past, have experimented with creating AI personas.

Creating a fake byline, if you will, where the organization might rationalize it by saying, “Well, we don’t want to credit this to a human journalist that didn’t actually write this.” So, they’ll credit it to something else, which maybe gives the impression that it was written by a human being.

I think most organizations now recognize that as a red line. You should not be creating the impression that something was entirely authored by a human when it was not—or that there is a person behind the story when it’s just an algorithm.

At the same time, when it comes to disclosure, [it’s getting harder to determine] when and what should be disclosed as AI becomes part of so many different [journalistic] activities. 

For example, if I’m a journalist, should I disclose that I used AI to help me come up with interview questions for a potential source? Does that rise to the level of meaningfully impacting a story to the extent that I need to disclose it? If I have to disclose that, do I also have to disclose the fact that I used generative AI to help come up with some of the potential story angles that I could cover? Do I need to disclose that I used it to help transcribe some of [my interviews]? 

[Journalists are] using it in so many different ways that if we say, “Well, you should disclose any use of AI,” then your AI disclosure is probably going to be almost as long as the story itself.

We also know from some of the research that when people see that AI has been used in any fashion, they immediately jump to the conclusion that it was used to basically report and write the entire story [without human oversight]—which is often not the case. 

I can use AI to help me narrow potential story angles [and suggest] potential questions to ask a source, and to help me do the transcription [while] still exercising my editorial judgment as a journalist about which of those angles to ultimately pursue, which questions to ultimately ask [and] which quotes to ultimately include in the story. [Journalists can use] AI in many ways that help [them] hopefully produce better journalism, but [without] making all of the decisions for [them].

I think when it comes to disclosure, we’re still figuring [things] out ethically, but I would just urge people to recognize that AI is already being used in so many ways—and that by using AI, [a journalist is] not delegating [their] entire editorial judgment to something else. They are still able to exercise editorial judgment when using AI.

In the book, you say that we still have the power to change the trajectory of how AI is used in journalism. How long do you think that window is open?

I don’t have a clear answer on when the window will close, but what I will say is that with each passing day, we are transferring more and more of our power and agency away—as citizens, as news consumers, and for news organizations for themselves—toward tech and AI companies that are having greater say over how this technology is being developed and implemented.

Each passing day of inaction makes it harder for us to re-orient the trajectory of AI development toward directions that I think would serve the public good better.

I don’t know when the window closes, but I would urge every one of us to be engaging in these conversations and to be demanding action from our elected leaders in terms of trying to promote more democratically oriented AI development. We can also vote with our wallets and what we choose to reward—from the search engines we go to, the apps we install on our phones, and by going directly to the information providers that we want to continue to provide us with information. 

There’s a lot that each one of us can do to be part of the solution, but I think we need to start taking action now. I think the story of inevitability that we so often hear with AI—the it is predestined to change the world and in preordained ways—that’s a story that we tell ourselves to avoid making the difficult and inconvenient choices that add friction to our lives in order to support what we want to see more of.

One of the main takeaways from the book is that we cannot reasonably expect news organizations to be the only entities that are part of the solution to the crisis in journalism that we’re facing right now in terms of reduced trust, reduced relevance, and so on. If we want to see more quality journalism, we need to be part of the solution ourselves—by rewarding quality news organizations by subscribing to them, by going directly to their content, and by asking our elected officials [and other leaders] to provide incentives, [be they] grants, tax breaks or direct regulation to make sure that quality media organizations and our media ecosystem are being adequately supported, so we have more quality information out there, and that there are disincentives for using AI in ways that exploit and ultimately pollute our information (networks).

Where can people find your book?

The book is available for free to anyone at journalismandai.com. A Kindle edition is also available now, if you want to support the authors. Paperback copies will become available in November in the United Kingdom and in January 2027 in the United States.