1. Let’s start with the nightmare scenario that keeps me up at night. A few years ago, when ChatGPT was still fresh and everyone was dazzled by its smooth answers and confident tone, a colleague asked me what my wildest fears were about generative AI. I didn’t hesitate. I described a world where bad actors could plant fake facts on slick, legitimate-looking websites and, in doing so, quietly reprogram the chatbots that millions of people now rely on for news and information. At the time, I was told by a tech CEO that I had a real problem dreaming up disasters. But here’s the thing: I wasn’t alone in having dark thoughts. The Guardian recently reported on a real-world example that looks exactly like the scenario I imagined. An Israeli-funded entity, posing as an American think tank, spent nine days flooding the internet with more than half a million words and over a hundred reports designed to push pro-Israel arguments. Seventy-three of those reports appeared in just two days. None of them had bylines. An advertising firm, paid nearly a million dollars through subcontracts from the Israeli government, created the whole operation. It’s the kind of revelation that makes you want to stop and ask: if this is happening on purpose, what else is already out there, quietly shaping what our chatbots believe and repeat? The answer, unfortunately, is probably a lot.
2. This kind of trickery is sometimes dismissed as “generative engine optimization” — you know, like search engine optimization, but for AI. That phrase makes it sound harmless, even clever, like some digital marketing tactic. But it’s not harmless. It’s not something you can shrug off and scroll past. When a chatbot ingests information that’s been published online, or sometimes even typed into its prompt box depending on the privacy agreement, that information doesn’t just float around for a moment and disappear. It becomes part of the model’s training data. It sinks down deep into the machine’s understanding of the world. And here’s the terrifying part: once that happens, the original sources and citations vanish. There’s no footnote anymore. There’s no way to fact-check it. It just becomes part of what the AI “knows,” indistinguishable from verified reporting and hard-won truth. That is a journalist’s worst nightmare. Nick Cleveland-Stout, a research associate at the Quincy Institute who wrote about the influence operation, put it bluntly: you won’t be able to fact-check it. Think about what that means. Every day, more people are using chatbots as their news source. They ask their AI assistant what’s happening in the world, and the AI answers with a blend of real reporting and synthetic slop, with no way for the user to tell the difference. We’re not just talking about a few biased articles showing up in a feed. We’re talking about the historical record itself being slowly rewritten, one poisoned chatbot response at a time.
3. Now comes the part where I might lose some of you. I have a controversial opinion, and I’m going to share it anyway. I believe that journalists and media scholars have to work together with technologists and technology companies. I know, I know. The thought makes many journalists cringe. There’s a long history of distrust between the newsroom and the tech world, and plenty of reasons for it. Tech companies have pilfered journalism at an unprecedented scale, using news articles to train their models without fair compensation or credit. Journalists are fighting for protections, for guidelines, for their work to be respected and paid for. That fight is absolutely necessary, and it should continue. But here’s the uncomfortable truth: our professions, the information ecosystem, and society as a whole are not going to thrive if we keep operating in separate silos, glaring at each other from a distance. The reality is that journalistic reporting has become a main character in a global war on truth. And in a war, you need allies, even allies who annoy you. As they say in the South, you’ve got to walk and chew bubble gum. You have to hold tech companies accountable while also figuring out how to work with them. You have to demand fair treatment while also learning how their systems operate so you can protect your own work. It’s not comfortable. But the alternative — refusing to engage, letting the machines run wild and the influence operations win — is far more dangerous.
4. So what does my advice look like in practice? Let’s start with the newsroom, because that’s where the fundamentals still matter most. The core work of journalists should not change. Reporting, investigating, writing, editing, and publishing the facts with rigor and ethical discipline — that remains the foundation. But news organizations also need to acknowledge that they are in a technological arms race. Their adversaries are savvy. They’re using AI to generate massive volumes of content, to game algorithms, to flood the zone with propaganda that looks like journalism. If journalists refuse to learn how AI distribution works, if they leave experimentation and mastery of what might be called “AI story optimization” to covert influence operations, then the rising number of people who use large language models for news will mostly encounter propagandized garbage instead of actual reporting. That’s not a future I want to accept. The good news is that journalists are not helpless. They have skills that machines don’t: judgment, context, accountability, and the ability to verify. They can learn how to make their reporting easily surfaced by AI systems without compromising their standards. They can partner with technologists who understand the mechanics of algorithms and search. They can experiment with new forms of distribution that meet people where they are, including the chat interfaces that are becoming so popular. The worst thing we can do is sit back and let the bad actors dominate the new terrain. We need journalists who understand AI to lead the way.
5. And what about the tech companies creating these robots? They have a responsibility too, and it’s one they’ve been avoiding far too long. Two years ago, I argued in CJR that the builders of large language models should work with journalists to fix how they handle breaking news. Since then, we’ve seen agreements between newsrooms and tech companies start to proliferate. That’s a sign of hope, and it suggests that this sort of partnership might be forming the basis of a new technological era, where legacy media organizations function as wire services for AI chatbots, providing trusted and timely information that can be easily surfaced algorithmically. But there are always human fingers tipping the scales. As LLMs determine which information on the internet is reliable enough to display, they rely on instructions — definitions and metrics that classify what counts as “news,” “trustworthy,” “credible,” or “true.” And I know from working at tech companies that these instructions are often written by people with no journalism background, no training, and, sometimes, no basic knowledge of journalistic essentials. That’s not an ideal situation. It’s deeply flawed, actually, and it’s easily solved with a little will and collaboration. There are companies like NewsGuard that have existed for years and already publish their rating and scoring criteria. AI companies could adopt those criteria or even ask their own chatbots to “read” and implement them. Academic centers, including the one I’m affiliated with at Columbia University’s Newmark Center, can and do consult with tech companies to develop better policies for determining whose voices should be amplified, especially in a world of “news influencers.” And sometimes, it’s just common sense. Even my journalism students could tell an AI company that if a website is publishing a torrent of articles with no bylines, you probably shouldn’t trust it.
6. Let me end with a kind of permission slip. I spent years working in tech, so I’m not naive enough to believe that companies have strong incentives to genuinely care about the truth. Profit margins, shareholder pressure, and the relentless race for dominance often push ethics to the side. But there are people inside those companies who do care, who are deeply concerned about the impact of large language models on our fragile information ecosystem. To those people, I say this: feel free to cite this article in your Slack message to your trust-and-safety policymakers. Seriously. Use me. An Ivy league professor gave you the green light. It may be uncomfortable for journalists and technology companies to work together. It may be awkward and frustrating and slow. But we need journalists who understand how AI works to partner with technologists who can fathom the work of journalism. We need everyone to put on their white hats and engineer solutions together. Because for all the recent talk of an impending AI apocalypse and humanity being annihilated in some catastrophic, amorphous, hypothetical manner, I think the more risky scenario is allowing these would-be god machines to continue their unfettered engagement in today’s information war. The real threat isn’t a Terminator-style robot uprising. It’s the quiet, steady corroding of truth. It’s a world where the historical record is so polluted with synthetic fakery that no one can remember what actually happened. It’s a world where the next generation grows up trusting machines that were quietly poisoned by people with agendas. We can prevent that. But only if we stop fighting each other and start fighting for the truth, together.