AI Is Rewriting Journalism From the Inside Out — And the Industry Isn’t Ready

Journalism has a new ghost in the machine. Artificial intelligence isn’t just assisting reporters anymore — it’s replacing them, rewriting their copy, and in some cases fabricating entire articles under the banners of trusted news organizations. A recent report from Communications of the ACM lays out the accelerating collision between AI tools and newsroom operations, and the picture it paints should concern anyone who cares about the integrity of public information.

The core issue is straightforward. Major publishers are deploying generative AI to produce and edit content at scale, often with minimal human oversight. Some are transparent about it. Many are not.

CNET made headlines in early 2023 when it was discovered the outlet had been quietly publishing AI-generated financial explainers riddled with errors. Futurism broke that story, and the fallout was immediate — corrections, retractions, and a credibility hit the brand is still absorbing. But CNET wasn’t an outlier. It was an early signal. Since then, outlets including Gannett, BuzzFeed, and Sports Illustrated have all experimented with or been caught using AI-generated content, sometimes attributed to fake human bylines. Sports Illustrated’s scandal, reported by Futurism, involved entirely fabricated author profiles complete with AI-generated headshots — a deception that cost the magazine’s publisher its licensing deal.

The ACM report traces how these incidents fit into a broader pattern. Newsrooms under financial pressure — and nearly all of them are — see generative AI as a way to maintain output while cutting staff. The math is seductive. One AI system can produce dozens of articles in the time it takes a single reporter to file one story. And the technology keeps improving. GPT-4, Claude, Gemini — each new model writes more fluently, hallucinates less frequently, and mimics human tone with increasing precision.

But fluency isn’t accuracy. That distinction matters enormously in journalism.

According to the ACM piece, AI-generated news content has introduced a new category of risk: plausible-sounding misinformation produced at industrial speed. Unlike traditional misinformation, which often originates from bad actors with clear agendas, AI-generated errors emerge from statistical pattern-matching. The models don’t understand what they’re writing. They predict the next likely token. When they get it wrong, the mistakes can look authoritative — formatted correctly, sourced-seeming, and written in the confident register readers associate with professional reporting. This makes AI errors harder to catch, both for editors and audiences.

The labor implications are stark. The Washington Post and other major outlets have conducted layoffs while simultaneously investing in AI tools. The Messenger, a digital news startup, shut down entirely in early 2024 after burning through $50 million — a collapse that underscored how fragile the economics of digital news have become. Into that vacuum, AI-generated content flows easily. Smaller outlets and local news operations, already hollowed out by years of declining ad revenue, are especially vulnerable. They lack the editorial infrastructure to vet AI output properly, and they’re the ones most tempted by the cost savings.

There’s a parallel battle over intellectual property. The New York Times sued OpenAI and Microsoft in December 2023, alleging that ChatGPT was trained on millions of Times articles without permission. The lawsuit, first reported by the Times itself, seeks billions in damages and could set precedent for how AI companies compensate — or don’t compensate — the publishers whose work feeds their models. Other outlets have taken a different approach. The Associated Press and Axel Springer struck licensing deals with OpenAI, essentially selling access to their archives. The split reflects a fundamental strategic disagreement: fight or cooperate.

Neither path is clean.

Licensing deals generate short-term revenue but risk training the systems that will eventually compete with the licensors. Lawsuits protect principle but take years to resolve, and the technology won’t wait. Meanwhile, AI-generated summaries in Google’s Search Generative Experience and similar tools are already siphoning traffic from publishers by answering user queries directly — no click-through required. The ACM report flags this as one of the most consequential long-term threats: not that AI will write bad journalism, but that it will make good journalism economically unsustainable by capturing the value of reporting without funding it.

So what’s the path forward? Some organizations are betting on transparency. The BBC and Reuters have published editorial guidelines specifying when and how AI can be used in their reporting. Labeling AI-assisted content is becoming more common, though standards vary wildly. The EU’s AI Act, set to take effect in phases through 2026, will require disclosure when content is AI-generated — a regulatory floor that doesn’t yet exist in the United States.

Industry groups are pushing for more. The Partnership on AI has proposed watermarking standards for synthetic content. The News/Media Alliance has lobbied Congress for legislation protecting publishers’ rights against AI training. Progress has been slow. Capitol Hill is still grappling with basic AI literacy, let alone the nuances of how large language models interact with copyright law and press freedom.

The tension at the heart of all this isn’t really about technology. It’s about trust. Journalism’s value proposition depends on human accountability — reporters who stand behind their work, editors who verify claims, institutions that issue corrections when they get things wrong. AI disrupts every link in that chain. Not because it’s malicious. Because it’s indifferent. It optimizes for output, not for truth.

And that indifference, applied at scale, could do more damage to public discourse than any single act of deliberate disinformation. The professionals building, deploying, and competing against these systems need to reckon with that — now, not after the next scandal.